Operating by John Brewton: Oper(AI)ting by John Brewton

How I Became Irrelevant. And How I Made Myself Relevant Again.

John Brewton — Thu, 02 Jul 2026 11:18:06 GMT

TL → DR

Two and a half years ago, a college intern on a consulting job I was working with zero hours in Excel and a $20 ChatGPT subscription completed work in 24 hours that I had blocked 25 to 35 hours to do myself, work built on skills I spent 20 years sharpening.
Being made irrelevant took 24 hours, and I never got a vote. Deciding to become relevant again took 30 seconds.
Fluency came from 1 method on repeat: Do. Fail. Learn. The skill that survived was not a skill. It was the decision to keep running the loop.

What happens when AI reprices 20 years of expertise?

The engagement was unglamorous. A consulting client needed to migrate to a new ERP system, which meant a mountain of dirty inventory data, the work no one wants. It was also the exact work I had spent 20 years building expertise in.

I had 1 person helping. A college intern, sharp, but with zero hours in Excel and zero in spreadsheets. Not a CS major. Not an engineer. Everything I knew how to do, he had never touched.

I handed him 3 things:

A laptop.
A ChatGPT subscription at $20 a month.
A pile of work I would not have time to finish.

The job required complex Excel workbooks, clean data, and advanced analysis. I had blocked 25 to 35 hours to do that work myself.

Subscribe now

How did an intern with no Excel skills finish 30 hours of expert work in 24 hours?

He came back in 24 hours. Most of it was done faster than I and cleaner than I. He also brought recommendations on where to take the analysis next and questions I had not thought to ask.

The method was simple. He never learned Excel. He asked ChatGPT to write code, dropped the code into Apps Script, and delivered everything I asked for and more. He never touched one of the spreadsheet skills I spent 20 years sharpening.

Sitting in that office, I said it out loud. I have become irrelevant. 20 years of hard skills, some of which I was very proud of, were repriced in an afternoon by a college intern with a $20 subscription.

What is the difference between being made irrelevant and becoming irrelevant?

I was made irrelevant. The repricing happened, and I never got a vote. Becoming irrelevant was up to me. Keep doing the old thing, and I would have finished the job myself, slower and prouder.

The repricing took 24 hours. The decision to do whatever I had to do to become relevant again took 30 seconds.

How do you become relevant again after AI reprices your skills?

The chain has 2 links. You become relevant by becoming fluent. You become fluent by running a loop, not by finishing a course.

Hard problems rarely yield to one attempt. They yield to a loop. You do. You fail. You learn. Then you run it again. Each turn carries the last correction forward. The failures get cheaper. The signal gets sharper.

How does the Do. Fail. Learn. Loop rebuild fluency in practice?

Do: The attempt, a specific bet put into contact with reality. I decided in that office and I tripled down. I pushed every workflow through these systems, smallest real attempt first, time-boxed. The chat tools were only the on-ramp. ChatGPT first, then Perplexity, then Claude, then Gemini. The first version was a probe, not the product.
Fail: The breakdown, reality returning a verdict I did not want. Prompts that produced garbage. Workflows that broke on contact with real client data. Automations I rebuilt 3 times before they held. Each failure was honest signal at contained cost. I named what broke, measured the gap, and kept the evidence. Failure was data, not identity. A cheap failure is a paid lesson.
Learn: The extraction, converting each verdict into a rule I could carry. Write the lesson down. Change 1 variable. Re-enter the loop. A lesson you cannot repeat is just a story. The correction is the asset.

Run with discipline, the turns compounded. What followed was more than automation. It was a retreat to first principles. Not the old workflow runs faster. A new operating model, with every process redrawn around what these systems can actually do.

Then the loop carried me past the chat tools entirely. I learned how databases were changing. I learned to ship AI-native applications. I learned to deploy agents through Cursor, Codex, and Claude Code, agents that do the work rather than answer questions about it. The work stopped stopping. I split jobs across machine instances running in parallel at 3 a.m., while I slept or worked on something else. The line that kept circling broke out.

What skill actually survives AI?

The skill that survived was not a skill. It was the decision to stay fluent. Staying fluent is refusing to exit the loop.

You will be made irrelevant. The shift is underway and you do not get a vote on that part. You do not have to become it. That part is a 30-second decision, and it is entirely yours. The decision has a method. Do. Fail. Learn. Run the loop until the breakout happens.

Read The Full 25 Here

Oper(AI)te is where you build the whole staff, not one skill: 8 sessions, 4 employees, and the one workflow that defines your job, built start to finish with Wessal Khader and me.

The summer cohort starts July 21, and for the next 48 hours, you can use code EARLY500 to receive a $500 discount on the regular price of $2500. Seats are selling fast; make sure to grab yours today!

Secure Your Seat Today

FREE DOWNLOAD

The Loop Log: the worksheet that runs the method. 1 page per turn, 10 turns. The bet, the verdict, the rule, the 1 variable changed. Print it or annotate it.

Download Your Loop Log For Free!

The 30-Day Fluency Sprint and the Failure Ledger arrive in the paid edition of this article on Saturday.

- j -

How do I build my first one?

You do not need to build all 25 to start. You need to build 1.

Become an Operating Founder. $99, one time.

What comes with it:

A live build session. We take one skill from this list and build it on screen, start to finish, yours to keep. You leave with a working employee and the method to build the next 24.
One year in my Skool community, with my Friday noon Q&A every week.
One 50-minute working session with me, one to one.

Become an Operating Founder Today!

About John

John Brewton documents the history and future of operating companies at Operating by John Brewton. He is a graduate of Harvard University and began his career as a PhD student in economics at the University of Chicago. After selling his family’s B2B industrial distribution company in 2021, he has been helping business owners, founders, and investors optimize their operations ever since. He is the founder of 6A East Partners, a research and advisory firm asking the question: What is the future of companies

Frequently asked questions

What is the Do. Fail. Learn. method? A 3-step loop for problems that do not yield to one attempt. You make the smallest real attempt, log the verdict reality returns, extract a rule you can carry, change 1 variable, and run it again. Each turn carries the last correction forward.

How is this different from just practicing with AI? Practice repeats the attempt. The loop repeats the correction. Without the written rule and the 1 changed variable, you run the same turn 10 times and call it experience.

How long does it take to become AI fluent? The decision takes 30 seconds. The first automation that holds takes most people 2 to 4 weeks of daily turns. I ran the loop from chat tools to deployed agents in under a year, working 60 to 90 minutes a day.

Do I need to know how to code? No. The intern who repriced me had zero hours in Excel and never learned it. He asked the model to write the code. The loop teaches you to direct the work, not to type it.

Which AI tool should I start with? Any of the big ones. Claude is strong for writing and voice. Perplexity is built for research. ChatGPT and Gemini both work. The tool matters less than the turn count.

How do I know when a lesson is real? Run the test in the Loop Log. A lesson you cannot repeat is just a story. If the rule changes your next attempt, it is an asset. If it only decorates the last one, re-enter the loop.

Will AI make my skills irrelevant? Parts of them, and you do not get a vote on that. The repricing of my Excel skills took 24 hours. What you control is the other side: whether you become irrelevant, which is a decision, and it has a method.

The 25 Most Important AI Jobs I Run For Myself & Build For My Clients Every Week

John Brewton — Tue, 30 Jun 2026 13:32:16 GMT

TL → DR

Most people run 2 jobs with AI. They ask a question and close the tab.
I run 25. Each one is a skill you can own, not a prompt you retype.
The list below is the whole staff, sorted into 5 groups.
The gap between 2 and 25 is the difference between using AI and running it.
11 of the 25 run as autonomous loops, several agents working together with no hand on the wheel.
You can build your first one this summer, live, in about 90 minutes.

What does an AI-fluent founder or operator actually do all day?

Most people are still using AI like a Google search box. They ask one thing, read the answer, and close the tab. And if the results aren’t strong enough, or seem not to make sense, they walk away and choose to do the work themselves.

That is their first mistake

That is 1 job, done by hand, every time, from scratch.

I run a team of +25 Agentic Employees every week. Not 25 prompts I retype. 25 agents that know my voice, my standards, my thinking, and my work. 25 team members who produce the same high-quality output every time and who adapt to the changing needs of my business each and every day

The gap is no longer between those who use AI and those who do not. The gap is who runs 25 jobs with it, while everyone else runs 2.

Here is the whole entry-level list of the work my team produces each week and the 25 most common first builds I engineer for my clients each month

Each one carries a tag.

Run means you operate it, one job at a time.
Scheduled means it runs on a trigger without you.
Loop means several agents run it end to end, in a loop, checking their own work, with no hand on the wheel.

What are the 25 things operators run with AI right now?

Writing in your voice

Draft a client email in 3 versions, one for the client, one for the team, one for the board, each in your tone, off one line. Run
Turn a 2-minute voice memo into a finished post in your own written voice. Run
Self-check every draft against your banned-phrase list before you read it. Run
Research a named prospect, write the cold outreach, then run the whole follow-up sequence on its own. Loop
Rewrite a stiff paragraph into your voice in under a minute. Run

Research that comes back sourced

Watch your calendar and deliver a sourced brief before every meeting, every claim cited, with no ask. Loop
Scan 3 rivals on a cadence, one agent each, and return the positioning gaps you can take. Loop
Turn 10 customer interviews into the 3 jobs they hire you for, one reader agent per transcript into one synthesizer. Loop
Fan out across a new market and return a sourced read in an afternoon, not a week. Loop
Pull a one-page memo on a person before every first call, triggered off the invite. Loop

Meetings that write themselves up

Turn rough bullets into a deck outline you would actually send. Run
End every call with the notes, the decisions, and the next steps already written, every time, automatically. Scheduled
Run the meeting pipeline, from transcript to notes to counterparty research to a relationship memo to a drafted email, on its own. Loop
Get the agenda and the pre-read from a 2-line goal. Run
Turn a long email thread into a summary and a recommended reply. Scheduled

Subscribe now

Money and deals

Normalize EBITDA on a founder’s spreadsheet and flag what will not survive diligence. Run
Read a deal room with one agent per section into one screen on whether the business is real. Loop
Turn a discovery call into a full proposal and a scope on its own. Loop
Build a 13-week cash view from raw numbers each week and name the weeks that get tight. Scheduled
Reconcile 2 months of statements and surface the duplicates and the gaps on a schedule. Scheduled

The work that is yours alone

Build the one weekly report that eats your Friday so it runs from your inputs each week. Scheduled
Stand up a brand voice profile and write everything through it. Run
Turn a research report into a finished carousel, research agents to writer to designer, charts and all. Loop
Triage your inbox every morning into needs-you, needs-a-reply, needs-nothing before you sit down. Loop
Build a custom skill for the one task no template fits. Run

Which of these run without me, completely?

Run, schedule, loop.

11 of the 25 reach the top rung.

Most teams stop at Run. They operate one skill at a time, by hand, and call that AI fluency. They are mistaken.

The ladder has 3 rungs. A skill you run becomes a job that runs on a schedule. A scheduled job becomes a loop that several agents run together, each one doing a piece, checking the work, and handing it to the next, until the result lands on your desk finished.

11 of the 25 above are at the top rung. A research brief that triggers off your calendar and writes itself before the meeting. A meeting pipeline that takes a transcript and returns a relationship memo and a drafted email with no prompt from you. A market read that fans 6 agents across the sources and comes back in the afternoon.

You are not prompting these. You build and schedule them once, and they run in perpetuity.

Caption: Four agents on one loop. The work moves without you.

That is the move most people have not seen yet. Not a faster search box. A staff of agents that runs the loop while you are in the room doing the work only you can do.

Why is this the summer to become AI-fluent?

Count how many of the 25 you run today.

If the number is 2, that is normal. It is also the problem.

Get Your AI-Fluency Score Today!

Waiting until the gap is loud is the costly move. By the time it is obvious, it has already cost you the deal, the raise, or the year.

Oper(AI)te is where you build the whole staff, not one skill: 8 sessions, 4 employees, and the one workflow that defines your job, built start to finish with Wessal Khader and me.

The summer cohort starts July 21, and for the next 48 hours, you can use code EARLY500 to receive a $500 discount on the regular price of $2500. Seats are selling fast; make sure to grab yours today!

Secure Your Seat Today!

How do I build my first one?

You do not need to build all 25 to start. You need to build 1.

Become an Operating Founder. $99, one time.

What comes with it:

A live build session. We take one skill from this list and build it on screen, start to finish, yours to keep. You leave with a working employee and the method to build the next 24.
One year in my Skool community, with my Friday noon Q&A every week.
One 50-minute working session with me, one to one.

Become An Operating Founder

What the $99 unlocks next:

Your seat in Oper(AI)te at $1,500 for the July 21 cohort, $1,000 off the $2,500 list.
A place in the affiliate program, 20% net on every seat you sell.

The $99 is the front door. The live build proves the method. Oper(AI)te is where you build the other 24, and the $1,500 rate closes with the July 21 cohort.

The founder rate is open for 14 days. When the 14 days end, the founder door closes. The list above is the menu. The first build is the key to the first item.

Become An Operating Founder

Frequently asked questions

What is the difference between using AI and running it? Using AI is one prompt at a time, from scratch, every time. Running it means you built a setup that already knows your voice and your work and does the same job the same way every time.

Can these really run on their own? 11 of the 25 can. Past the single skill, a job runs on a schedule, then several agents run it as a loop, each doing a piece and checking the work. You set it once and it runs without you.

Do I need to know how to code? No. You answer questions and paste in samples. The build does the rest.

Which AI app should I use? Any of the big ones works. Claude is strong for writing and voice. Perplexity is built for research. ChatGPT and Gemini both work. Pick one and start.

How long does the first one take? About 90 minutes in the live build session. Then you run drafts to make it sharp.

What does $99 get me? Lifetime access to the Operating Founder community, the weekly Friday office hours, a new weekly AI fluency session through the summer, one live build session, the right to join Oper(AI)te at $1,500 for the July 21 cohort, and a place in the affiliate program at $300 a seat.

What happens after the 14 days? The founder door closes. The $99 rate is sold only inside the 14-day window.

Is this safe with my private writing? Your profile lives in your own account. You read every file before you turn it on.

About John

AI Fluency, Defined: The 5 Moves That Replace Prompt Lists

John Brewton — Sat, 27 Jun 2026 11:32:21 GMT

TL → DR

AI fluency is the capacity to see the whole workflow and build it yourself. Not knowing a tool. Operating one.
Prompt lists are a phrasebook. They run out. Fluency is a grammar. It does not.
The discipline is 5 moves: Frame, Assign, Brief, Inspect, and Iterate then persist.
You drill them on small prompts until they are automatic. Then the same 5 moves run multi-agent systems.
Paid subscribers get the tool: the Rudiments skill for Claude and ChatGPT, plus a one-page chart.

Subscribe now

I live in a bilingual household. My wife is from Mexico; Spanish is our first language at home, and our two-and-a-half-year-old daughter is becoming bilingual day one. She moves between Spanish and English as if it’s all one language. She produces sentences no one taught her. It’s one of the most beautiful experiences of my life.

I studied Spanish for 4 years in high school and Chinese for 4 years in college. 8 years of classes, and I could not have held a conversation in either. I never became fluent. I was memorizing lists for the test and forgetting them by summer. Since meeting my wife, I have worked hard to become bilingual. I’m not quite there yet, but getting very close. What is making me fluent now is my household, my wife, our daughter, our friends, and my extended family. I speak more Spanish than English in my personal life each day.

That is how most people should learn AI, and almost no one does. They collect prompts instead. “10 prompts for marketers.” “The 50 best ChatGPT prompts.” They are building a phrasebook, and they will struggle to produce high quality outputs the first time the work is not on the list.

AI fluency is the capacity to see a workflow whole and build it yourself, which is the one thing software has spent your entire career doing for you.

Nobody has defined AI fluency, so the word floats. People feel behind without knowing what they are behind on. Companies buy licenses and wait for a result that never arrives.

The gap is not access to the tools.

The gap is a definition nobody wrote down or thought to calrify.

What does it actually mean to be fluent?

Linguists test it with 3 markers:

A fluent speaker stops translating in their head.
They route around the words they do not know instead of freezing.
They create instead of reciting.

My daughter is 2 and already does the third one. Most adults with a language app never clear the first.

The same 3 markers tell you a person is fluent with AI. They stop reaching for a saved prompt. They route around what the model cannot do instead of giving up. They compose a workflow for a problem no one has templated.

New to Operating? You Might Appreciate These Popular Articles

Why is working with AI different from using software?

For 40 years, enterprise software has made this skill unnecessary. The ERP told you which screen to open. The CRM told you what to enter and when. The workflow was built into the software, written by a vendor who had already anticipated your situation. You pointed at the phrase.

That arrangement ended in 1 step. The model does not hand you a directional form on a screen; it hands you a monumentally flexible canvas through which you can communicate to build anything, literally.

The cost of building a workflow fell to a sentence you can type. That changed what a tool is. Which changed what a worker does. This is now changing what it means to be a worker. Most people are still asking which prompt to use.

We have spent our careers doing work that computers could not do, yet. Now the computer does the work and asks us to compose and direct its intended output. That is not a smaller job. It is the work of an Operator, and very few companies, much less individuals, have been trained for this task.

Can you become AI fluent without a technical background?

I trained as a drummer at the Berklee College of Music for a year right after high school. I obsessively drilled rudiments, single strokes, double strokes, paradiddles, slowly and cleanly, thousands of times, until my hands moved without thought.

When the rock-and-roll dream didn’t work out, I became an equally obsessive student, ultimately graduating from Harvard with a degree in Economics. The disciplines and habits that helped me master an instrument were the same tools that helped me succeed in academia.

I did it again in business. I spent a decade at a B2B industrial distributor, from the sourcing desk to the chief executive role, and grew it from $27.5M to $75M before selling in 2021. The work computers could not do yet was the work I did by hand for 10 years.

I am doing it a third time now. I am building AI-first, enterprise-grade software using tools Attio, Claude, Supabase, and Resend. I am not an engineer. I am an Operator, a buildeer, a designer, and a founder.

The discipline that produced all 3 outcomes across my life transfers and compounds with each successive iteration.

For AI Fluency, this skill-building repetitive loop has a clear shape, I call it the Five Moves. It is rudimentary, and meant to be practiced over and over again

You do not become AI Fluent by reading about reps. You achieve this new reality by building this capability across your real work, alongside colleagues who are doing the same, and by learning from leaders who have already achieved the outcome.

That is what I built with Wessal Khader. It is called Oper(AI)te, and it runs this summer. 8 live sessions, no theory, and no replays to watch alone. You build a working staff of AI employees that does your real work, across Claude, ChatGPT, Gemini, and Perplexity. Beginner or advanced. No coding ability required. You leave with a team of agentic AI-employees running 24/7.

The summer cohort starts July 21, and seats are limited.

Take Your Seat @ Oper(AI)te!

Build it with us this summer to become AI-Fluent by the time Q3 ends and Q4 starts!

What are the 5 moves of AI fluency?

The five never change. Only the size of the task changes. You drill them on a one-line prompt the way a drummer drills a single-stroke roll, then the same 5 moves run a multi-agent system. Start small on purpose. A weak prompt costs nothing, and you can run the loop 40 times in a morning on real work.

Frame: One question: what is the actual outcome, and what does done look like? Small: “summarize this” becomes “pull the 3 decisions and who owns each.” Large: map where the workflow begins, ends, and every touchpoint between. Skip it and you accept the first plausible thing the model returns, because you never defined better.
Assign: One question: who does each part, you, the model, a tool, or an agent? Small: decide whether the model is thinking or just formatting a decision you already made. Large: the model’s job, a tool’s job, a sub-agent’s job, your judgment. Skip it and you hand the model work it cannot do, then blame it for guessing.
Brief: One question: what do you know that the model does not, and what must it not do? Small: paste the constraint, the audience, and one example of good. Large: persistent context, data sources, guardrails. Skip it and the answer comes back fluent, confident, and generic, because you described a generic situation.
Inspect: One question: where is this wrong, thin, or bluffing, measured against move one? Small: read the output as an operator checking work, not a customer accepting a delivery. Large: test points, spot checks, the places a workflow breaks quietly. Skip it and fluent-sounding wrongness ships. This is the most expensive failure in the discipline.
Iterate, then persist: One question: what one change improves this, and how do you keep what works? Small: change one variable, run again, notice what changed. Large: save what worked as a prompt, a project, a skill, or an agent, so it runs again without you. Skip it and every task starts from zero, and nothing compounds.

How do you practice AI fluency?

The cost of a mistake on a 2-line prompt is limited. The rep count is everything. Run the loop 40 times a day on real work, and the moves leave your conscious attention. By the time the task is a system with sub-agents across 4 tools, the moves are already in your hands, and your attention is free for the one thing that does not automate: judgment.

The rudiments are rudimentary. That is the point. Mastered, they become the foundation for building things that run without you.

Is there a tool that enables the 5 moves?

I built a skill that runs you through the 5 moves while you work. It is called Rudiments. When you sit down and say “help me draft the board update” or “build me a lead-routing flow,” it does not lurch into the task. It asks you the moves, calibrated to the size of the job, and it shows you the difference between your loose ask and the sharp one, so you learn the move and not just the answer. It is built to make itself unnecessary.

Here is what it does with every task you bring it:

Sizes the job. A quick rep for a one-line prompt, a full rep for real work, a build rep for a multi-agent system.
Walks you through the 5 moves as questions: Frame, Assign, Brief, Inspect, and Iterate then persist.
Shows you your loose ask next to the sharp one, so the rep teaches you the move.
Tracks the moves you skip and tells you which one to drill next.
Saves what works, so the rep compounds instead of resetting to zero.

That is the loop and the rep that will help you get one step closer to becoming AI Fluent.

You do not need to write code. You do not need a developer. You need Claude Pro at $20 a month or a Premium ChatGPT account, and 60 minutes of real work to drill against.

Paid subscribers get 3 things, attached below:

The Rudiments skill for Claude, installed in under a minute.
The same discipline written as instructions for a ChatGPT, so it runs wherever you work.
The one-page chart, print-ready.

Become a paid subscriber to download the Claude skill and the directions for how to build the same capability in ChatGPT after the frequently asked questions section below.

Subscribe now

Frequently Asked Questions

What is AI fluency? AI fluency is the capacity to see the whole workflow and build it yourself. It is not knowing a tool. It is operating one. The test is the same as for language. You stop reaching for a saved prompt, you route around what the model cannot do, and you compose a workflow for a problem no one templated.

What is the difference between AI fluency and prompt engineering? Prompt engineering collects lines that work. AI fluency builds the workflow that those lines sit inside. A prompt list is a phrasebook, and it runs out. The 5 moves are grammatical, and they do not.

Do you need to know how to code to become AI fluent? No. I am building an AI-native CRM for a public company, and no one trained me to write software. You compose the workflow in plain language and let the model and tools handle their parts. You need judgment, not a computer science degree.

What are the 5 moves? Frame, Assign, Brief, Inspect, and Iterate then persist. Frame the outcome. Assign each part to you, the model, a tool, or an agent. Brief the context and the guardrails. Inspect the output against the outcome. Iterate one change, then save what works so it runs again without you.

How long does it take to become AI fluent? Faster than you expect, because you drill on small tasks with a high rep count. Run the 5 moves 40 times in a morning on real work, and the moves start to leave your conscious attention in weeks, not years.

Should I use Claude or ChatGPT to practice? Either. The discipline is identical on both. Claude Pro is $20 a month, and ChatGPT has a free tier. The Rudiments skill ships for Claude, and the same discipline is written as instructions for a ChatGPT Custom GPT.

What is the Rudiments skill? It is a tool that runs you through the 5 moves while you work, calibrated to the size of the task, and shows you the difference between your loose ask and the sharp one so you learn the move. It is built to make itself unnecessary. Paid subscribers get it for Claude and ChatGPT, plus the one-page chart.

Download the first skill you need to become AI Fluent below.

How to Build Your Own AI Employees: A Simple 2026 Guide

John Brewton — Sun, 21 Jun 2026 00:17:56 GMT

What is this article about? (TL → DR)

Most people use AI like a search box. They ask one thing, then close the tab.
A few people do more. They build AI “employees” that do the same work every day.
An AI employee is a setup you create once. It learns your voice, your job, and your rules.
Four are worth building first: one for writing, one for research, one for meetings, and one for your own job.
You can build the first one this week. No code. The free kits and a free live class are at the bottom.

Want to build one live with me, for free?

Register for my free Masterclass on June 16th, where I will teach you how to build your own AI Employee that runs your calendar and inbox! Register Here

A free 60-minute live class with me and . You come in with a full calendar and a packed inbox. You leave with an AI employee that runs both, and the way to build the next 10. No code. No replay.

Reserve Your Free Seat

What is the difference between using AI and operating it?

Using AI is simple. You open it, ask a question, and close it. The answer sounds flat. The tool knows nothing about you.

Operating AI is different. You build something that stays. You teach it your voice, your job, and your rules. Then you run it like a teammate who already knows your standards.

Here is a typical Monday for a person who built their staff:

A big slide deck is due on Wednesday. It gets drafted in an hour, not over two nights. Three calls run back-to-back.
Each one ends with the notes, the choices, and the next steps already written.
A question that used to eat up an entire afternoon comes back with a sourced answer before lunch.

That is not a dream. That is what the right setup does.

What is an AI employee?

An AI employee is not a prompt you copy from Reddit or copy and paste from a LinkedIn post. It is a setup built for one job inside your team and workflows. You engineer it one time. It runs the same way every time. You adjust and tweak, but do not ever need to rebuild.

These are the four most worth building first:

The Voice Employee writes like you. You train it on your real emails and posts. It can write to a client, to your team, or to your boss, each in the right tone. It checks its own work before you see it. Done right, no one can tell you used it.
The Research Employee gets you ready. You give it the questions you ask every week. It pulls real facts, sorts them, and hands you a short report with sources. It preps you before any meeting that matters.
The Meetings Employee kills the blank slide. You give it your rough notes. It hands back a clean deck you would actually send. It writes the agenda. It turns your call notes into a summary. Every meeting ends with the work already done.
The Role-Specific Employee is yours alone. You have one task that eats hours it should not. A report. A proposal. A weekly grind. No template fits it. You build this one around your real job.

Connect With Me On LinkedIn

Connect on LinkedIn

Why do most people get AI wrong?

Most people stay stuck for one reason. They think of AI as a tool. Tools are passive. You pick them up, use them, and put them down.

An employee is different. It carries what it knows. It knows your standards. It knows what you would say no to. You do not explain yourself again every time.

Change that one idea, from tool to employee, and everything changes. How you build it. How you talk to it. What you expect to receive back.

Consulting Perspective: This is one of the biggest mental hurdles I see inside my clients’ companies. Individuals struggle to provide enough direction, detail, and specificity when they’re prompting one of the models to perform a taks, or within the build process for the engineering of their agent.
When you make the jump to looking at the AI as a new hire, a talented, smart analyst, and begin providing said analyst with all of the context, reference materials, and clear direction that you would to any team member you were asking to produce high-quality work product, that is when the quality of the results, the outcomes, and the strength of the builds you are engineering, dramatically improve.

The primary apps are not hard to use. Claude is strong in long writing and voice. Perplexity is built for research with sources. ChatGPT does a little of everything very well and is currently the cat’s meow for image generation. Gemini sits within Google, where many people already work, and has a unique capability to help with management and team-dynamics queries. You do not need to master one. You need to know which one to grab, and why.

Are most people actually good at AI?

Here is the hard part. Most people in your meetings are faking it. They put “AI” in their slides. They say they are “trying the tools.” They paste prompts into free accounts and take screenshots of the results.

The gap is getting larger. You can see it in who writes the clearest report the fastest. Who shows up to the meeting with the work done? Who gets more done without longer hours? That person, most likely, quietly built their staff.

The most costly move right now is to wait until it is obvious. By the time the gap is loud, it has already cost you.

How do I start building my AI staff?

You can build your staff this summer. Do it before the calendar fills and the year resets. Do it before a skill gap turns into a career gap.

Start with the Voice Employee. Train it on 10 to 15 real things you have written. Tell it what you never say. Tell it how you sound when you are rushed. Tell it what “clean but not stiff” means for you. Run 30 drafts before passing judgment.

What are the AI staff kits?

I built the four employees as kits so you do not have to start from scratch. One kit per employee: Voice, Research, Meetings, and Role-Specific.

A kit is a skill file. It holds the steps that run the job, the notes that carry your standards, and a template for the output. That is the gap between a prompt you retype and an employee who already knows the work.

Each kit has 2 modes. BUILD sets up your profile the first time. Your profile is the file that holds what the kit learned about you. OPERATE runs the job at your standard after that, and checks its own work before you see it.

How do I install and run the kits?

First, install it. If your app takes skill files, load the kit. If your app uses projects instead, paste the steps into a project and add the notes as files.

Second, run BUILD one time. Answer the questions. For Voice, that means 10 to 15 writing samples. For Research, your weekly questions. For Meetings, your formats. For Role-Specific, the one task worth it. This makes your profile. Keep it. It is the part that matters most.

Third, run OPERATE from then on. Give it the input. Take the output. Fix what missed. Save the fix back into the profile. It gets faster each time.

Start with the Voice Employee. Build it this week. Run 30 drafts before you judge it. Then add the next one. Each one makes the next build faster, because the work you did once carries over.

That is the whole setup.

Why does my AI output still sound generic?

The build misses in 2 ways. Both are quick to fix.

The skill does not run. The app only uses a skill when the description tells it when to. If nothing happens, just say it: “Use my Voice Employee to draft this.” Then check two switches. The skill and code tools are turned on in settings. One of those is almost always the cause.

The output still sounds robotic. That is a thin profile, not a broken tool. Add 5 more writing samples. Use the ones you wrote in a hurry. Make the “never say” list longer. Name 3 phrases you would never use. Run 10 more drafts. It gets sharp from your fixes, not from the first try.

When the output misses, fix the profile, not the prompt. The prompt stays the same. The profile is the part that learns.

Want me to build it with you?

We take one problem in your business and finish it across four sessions. You pick the problem: your pricing, your brand, your content plan, or one issue inside a company you already run. We name it, we plan it, we build the machine inside Claude or ChatGPT, and we put it to work on real tasks.

It normally sells at $999. The first 10 seats are $499, half off. Each seat includes a full year in the Operating Founders community in Skool and weekly Friday office hours. You can run your four sessions any time through December 31, 2026.

Claim Your Seat

Frequently asked questions

Do I need to know how to code?

No. You do not write code. You answer questions and paste in samples. The kit does the rest.

Which AI app should I use?

Any of the big ones works. Claude is great for writing and voice. Perplexity is great for research. ChatGPT and Gemini both work too. Pick one and start.

How long does it take to build the first one?

About an hour to set up the Voice Employee. Then run 30 drafts to make it sharp. You can do it in a week of short sessions.

Is this safe with my private writing?

Your profile lives in your own account. Only load kits from people you trust. Read the files before you turn them on.

What if the output still sounds robotic?

Add more writing samples and a longer “never say” list. Fix the profile, not the prompt. It learns from your fixes.

How much does it cost?

The kits are free. A paid AI plan runs about $20 a month. A free plan is enough to start.

Do I have to build all four at once?

No. Build the Voice Employee first. Add the others when you are ready. Each one makes the next faster.

What is the difference between a prompt and a kit?

A prompt is one message you retype. A kit is a saved setup that already knows your job and runs the same way every time.

Where do I get the kits?

Have a look at the Field Guide here and then become a paid subscriber to access all of the skill kits and files below.

Become a paid subscriber for as little as $7.99 per month and access them below…

Subscribe now

The Operating Project: Most Coaching Ends With Notes. This Ends With a Machine.

John Brewton — Mon, 15 Jun 2026 23:26:19 GMT

Most coaching ends the same way. You leave with notes, a few good ideas, and the quiet knowledge that in three weeks the notes will be cold and the ideas will still be ideas. The work that was supposed to change did not change.

You paid for advice. But you needed a system.

I operated, optimized, and scaled a B2B distribution company from $27.5M to $75M and sold it in August of 2021. I have spent more than 400 hours this year coaching founders and operators. I’ve learned so much from the incredilbe +100 founders, entrpreneurs, operators and professional I’ve been fortunate to work with. Many of these amazing people have become my friends and collaborators and for that I am grateful.

I wanted to take all that I’ve learned from these individual engagements and turn them into a product and system to drive outcomes for the Operators and Builders I’m proud to call my clients.

What I came up with is the Operating Project.

Claim Your Seat!

What it is

The Operating Project takes one problem in your business and works it to a finished result across four sessions. You pick the problem. I bring the structure, the worksheets, and the build. The outcome is defined before we start, and the project closes when the finished thing exists.

This is not open-ended coaching. It is not a standing retainer. It is one operating problem, scoped to one outcome, closed.

Pick one problem

You choose the project. Four kinds of work come up most often.

A business model. Your pricing, your offer design, and the economics of what you actually sell.
A brand identity. The system that makes your work recognizable across every surface, instead of a new look every time you post.
A content strategy. The plan and the machine that produces it on a schedule, instead of the weekly scramble.
A matter inside a larger business. One specific problem in a company you have already built and need to fix without rebuilding the whole thing.

One problem. One outcome. One finished artifact you can use.

Claim Your Seat!

The four sessions

The work runs in four moves.

First, we diagnose. We name the problem in plain terms, set the outcome you want to hold at the end, and define the finished artifact. We draw the scope and what sits outside it.
Second, we design. We build the actual plan. Pricing models, offer maps, positioning frameworks, content calendars, the worksheet that fits your project. You build them with me and you keep them.
Third, we build the machine. This is the part nobody else includes. We stand up a working system inside Claude or ChatGPT that produces the work you just designed. The skills, the projects, the workflow. You own it. It lives in your account.
Fourth, we install. We run the system on a live piece of your business, confirm it holds, and define done.

The sessions are not timed. Some run thirty minutes. Some run longer. The unit of measure is progress, not a clock.

Claim Your Seat!

A Plan Tells You What To Do. A Machine Does It For you.

Most advice dies because it stays a document. A document needs you to remember it, apply it, and keep applying it after the energy of the session fades. A machine does not.

That is the difference here. You leave with a system that runs after I am gone. I am in these tools every day, building the systems I write about. The build is the same work I do for myself, handed to you and tuned to your business.

What People Are Saying About John

“John doesn’t give you advice. He gives you clarity.” Matt Burton, founder of Savant-AI.

“His perspective on AI is practical and lived-in. He is not giving abstract advice from the sidelines. He is in the tools himself, testing workflows and building systems.” Arvita Tripati, founder of Vahana Labs.

“A savant when it comes to operations. I trust his opinions and respect his instincts when it comes to making my business a success.” Laura LeBleu, founder of Geezer Magazine.

Claim Your Seat!

The Terms

The Operating Project sells at $999. I am opening ten founding seats at $499, half the price, for the first ten people who take one.

When they are gone, the founding rate is gone. Ten is the most I can deliver well alongside the work I already carry, so the cap is real.

Each seat includes a full year of VIP access to the Operating Founders community and the weekly Friday office hours.

You can run your four sessions any time through December 31, 2026. Lock the seat now, start when you are ready.

Take one of the ten before they’re gone!

Pick the problem. I will help you solve for it, and build the machine that keeps it solved.

Claim Your Seat!

- John -

About John

The AI-Fluent Consultant: How To Scale Your Consulting Company by Becoming AI-Fluent

John Brewton — Sat, 13 Jun 2026 11:31:58 GMT

By John Brewton · Founder · Operating by John Brewton

Start Here To Become An AI-Fluent Consultant

Are you a solo consultant or professional services provider? Whether you specialize in financial analysis, acquisition strategies, accounting, or brand marketing and sales, this article is tailored for you. In today's AI-driven landscape, consultants and service providers can expand their businesses faster and more efficiently than ever before. If this resonates with your work, you won't want to miss these transformative insights.

Let's dive in.

What once took 40 hours for a client audit now takes just 4. At $200 per hour, that efficiency comes at a cost of $7,200 in lost billable time. The billable hour penalizes consultants who create more efficient systems.

Over the past six months, I've tested the full Claude and Perplexity stack against the real-world workflows of a solo advisory practice. This article draws from 24 primary sources and 12 academic and institutional studies, all linked inline and organized in the appendix. The conclusion is clear: integrating AI into your current practice can boost efficiency by 20 to 30 percent, but redesigning your operating model around AI enables the productivity of a 3-to-5-person firm as a solo consultant.

TL;DR

AI-fluent means rebuilt, not retrofitted. Speeding up old workflows yields 20 to 30 percent gains. Redesigning the operating model around AI yields 5 to 10x output on research-heavy work, per Digital Applied’s micro-consulting benchmarks.
The stack is 2 tools with different jobs. Perplexity is the intelligence layer; it reads the live world and returns cited research. Claude is the execution layer; it holds your firm’s IP and produces the deliverables. Together, they close a loop: verified intelligence in, client-ready output out.
The evidence is no longer anecdotal. A field experiment with 758 BCG consultants found that AI users completed 12.2 percent more tasks, 25.1 percent faster, and at 40 percent higher quality on tasks within the model’s competence, per the Harvard Business School jagged frontier study.
Demand confirms the shift. AI-augmented freelance consulting skills grew 109 percent year over year in Upwork’s 2026 In-Demand Skills report.
Pricing is the play that captures the value. When delivery compresses, hourly billing becomes a liability. Outcome-based pricing converts the time savings into margin and raises practice exit multiples from 4x toward 6x EBITDA, per DEMG’s analysis.
The risks are manageable, not optional. Verification, data handling, and billing ethics decide whether the system compounds or collapses.

This week’s article builds one thing: a working AI operating system across Claude and Perplexity. The Vault hands you every asset to build it.

Twelve assets: Six reference and instruction files. Five working skills. One Companion Field Guide that compiles it all into a 24-page PDF.

The five skills do the work: ICP scoring, meeting intelligence, proposal generation, Five Forces, and Jobs To Be Done.

Free subscribers get the thinking. Paid subscribers get the build.

Subscribe and download the Vault.

Become A Paid Subscriber Today

What does “AI-fluent consultant” actually mean?

The phrase has been stretched to cover everything from a ChatGPT tab to full process reinvention. The useful definition is narrow. An AI-fluent practice is one in which the operating model is designed around AI from the start, rather than retrofitted onto a legacy service-delivery structure.

The distinction shows up in the economics, not the tooling. BCG’s 2025 study of 1,250+ companies found that only 5 percent achieve AI value at scale, while 60 percent report no material value despite real investment. The gap is not access to models. Everyone has the same models. The gap is in the operating model design.

The same split is restructuring the big firms. HBR documented in September 2025 that AI is collapsing the consulting pyramid into an obelisk, automating the research and modeling that justified thousands of junior billable hours. The FT reported in May 2026 that McKinsey is overhauling partner pay under Project Acorn because roughly a quarter of global fees are now tied to outcomes rather than hours. The Economist asked in June 2025 who needs Accenture in the age of AI, after the firm had shed roughly $60 billion in market value since its February 2025 peak.

The pyramid is breaking at the top of the market. That is the opening for the solo operator at the bottom of the cost curve.

What is the best AI stack for a solo consultant?

Claude is the execution environment: It holds your firm’s IP, frameworks, voice, and processes in Projects and Skills. It produces the deliverables clients pay for through Code, Artifacts, and MCP-connected tools.

Perplexity is the intelligence layer: it ingests the live world, web sources, internal files, competitor signals, and earnings transcripts, and routes these signals through specialized models to produce verified, cited research faster than any human team.

The Loop: Perplexity generates the intelligence. Claude shapes it into client-ready output. Your judgment sits between them as the quality gate.

Perplexity intelligence layer (Deep Research, Tasks, Spaces, Computer) feeding the operator gate, feeding the Claude execution layer (Projects, Skills, Artifacts, MCP), feeding client deliverables, with the feedback loop returning to intelligence.

Subscribe now

What are the 10 highest-leverage AI plays for a consulting practice?

The full research report identifies 10, ordered roughly by implementation complexity.

Deep Research for first-pass analysis: A 30-page cited report in under 4 minutes replaces 20 to 40 hours of engagement research.
Always-on market monitoring: Scheduled intelligence replaces 3 to 5 hours a week of manual scanning across client verticals.
Claude Projects as the firm’s operating brain: Your frameworks, voice, and pricing persist across every conversation.
Claude Skills for repeatable frameworks: Porter’s Five Forces, ICP scoring, JTBD synthesis, encoded once, run in minutes.
Discovery-to-proposal automation: 48 to 72 hours compressed to under 2.
First-draft deliverables via Artifacts: Interactive memos, 11-tab financial models, working mini-apps.
CRM intelligence via Attio MCP: Research, record creation, and outreach drafting in 1 conversational loop.
Perplexity Spaces as engagement hubs: Internal files and live web searched together.
The content engine: 1 newsletter becomes 10+ platform assets, client work becomes thought leadership.
Outcome-based pricing: The structural play that converts the first 9 into revenue.

Subscribe now

How do consultants use AI for research without getting burned?

Perplexity Deep Research performs dozens of searches, reads hundreds of sources, and synthesizes a cited report in 2 to 4 minutes. It scores 93.9 percent on SimpleQA. The scope change matters more than the speed change. A human analyst carefully reads 20 to 30 sources. Deep Research reads hundreds, which changes the statistical odds of surfacing the non-obvious insight.

The discipline that makes it safe runs in 3 steps. First, run the research inside a Perplexity Space scoped to the engagement, with source instructions set to the client’s vertical. Second, export the cited report into the engagement’s Claude Project as reference material. Third, spot-check 5 to 10 citations per report before anything enters a client deliverable. That takes 15 minutes, compared to hours of independent verification.

The productivity evidence behind this play is peer-reviewed. Noy and Zhang’s MIT experiment, published in Science, found that access to ChatGPT reduced professional writing time by 40 percent and improved quality by 18 percent. Ethan Mollick documents a Danish study in which AI halved the working time on 41 percent of tasks. The Stanford and NBER call-center study found a 14 percent average productivity gain, with 34 percent for novices. Research-intensive consulting tasks compress harder than any of these because parallel search structurally beats sequential browsing.

Do you want to start becoming AI-Fluent?

Register for my free Masterclass on June 16th, where I will teach you how to build your own AI Employee that runs your calendar and inbox! Register Here

Can AI monitor my markets while I sleep?

Yes, and this is the play most consultants skip. Perplexity Computer, launched February 25, 2026, runs asynchronous multi-step work across 19+ frontier models. Perplexity Tasks schedule recurring Deep Research queries delivered to your inbox.

The standing configuration for a 5-client practice: competitor funding and hiring scans weekly, contract awards from SAM.gov daily or weekly, regulatory changes weekly, client account news daily, earnings synthesis quarterly. Karo Zieminski documents 5 weekly workflows running in Computer, including the warning that matters: the May 2026 default orchestrator switch changed long-running task behavior, so pin specific models for production workflows.

The gain is 3 to 5 hours a week. The bigger gain is never missing the competitor funding round or policy change that should have triggered proactive client outreach.

Subscribe now

How do I make Claude actually remember my firm?

Claude Projects persist documents and instructions across all conversations within them. The architecture that works: 1 Project per client engagement plus 3 standing firm-IP Projects.

Firm Voice and Methodology holds your style guide, frameworks, and anonymized past deliverables. Build this one first. Proposal Factory holds pricing structures, SOW templates, and scoping criteria. Research Synthesis Hub holds recurring frameworks and sector source lists.

Claude Skills are the action layer on top. A Project configures behavior. A Skill executes work: it takes inputs, runs multi-step logic, and produces a structured output. The canonical consulting Skills are a Five Forces analyzer, an ICP scoring engine, a JTBD interview synthesizer, and a proposal generator. One practitioner documented cutting proposal time from 2 hours to 90 seconds with the Skill pattern. A framework application that took 4 to 6 analyst-hours to build runs in 10 to 30 minutes. Across 5 engagements a month, that is 20-30 hours recovered.

The design principle: a Skill produces a specific, reviewable output, not a conversation.

How do I automate the discovery-to-proposal pipeline?

This is the most commercially sensitive workflow in any practice, the gap between qualified interest and signed engagement. Speed here moves close rates. Stack.expert’s pricing analysis notes AI-augmented consultants quote 2-day turnarounds where traditional firms quote 4 to 6 weeks, and clients pay premium rates for the speed.

The pipeline runs in 5 steps. A meeting notetaker routes the discovery transcript to the engagement Project via MCP. The Proposal Skill extracts the stated problem, the unstated needs, success criteria, and decision process into a 1-page brief. The Skill matches extracted needs against your offerings file and recommends 1 to 3 packages with pricing rationale. Claude drafts the full SOW in your template and brand format. A final Perplexity pass adds 1 or 2 fresh data points on the prospect to the executive summary.

Subscribe now

Can AI build the actual client deliverables?

The deliverable layer is where the stack stops being a research assistant and starts being a production system.

Claude Artifacts produce interactive outputs: strategy memos with embedded charts, working ICP scoring tools, market sizing calculators, competitive positioning matrices a client can explore. As of April 2026, Live Artifacts connect to external tools via MCP and refresh in real time. Anthropic’s Financial Services solution ships pre-built plugins for financial modeling. One analyst built an 11-tab financial model in 10 minutes including sensitivity analysis. Perplexity Labs is the research-native equivalent, producing dashboards and reports grounded in cited web sources.

On the pipeline side, Attio MCP gives Claude OAuth-authenticated access to your CRM. The flow Attio documents: run deep research on a target in Perplexity, then instruct Claude to create the company record, add contacts, log the research as a note, and draft personalized outreach referencing a specific finding, all in 1 conversation. Estimate 3 to 5 hours a week saved across a 20-to-30-account pipeline, on engagements worth $15,000 to $150,000 each.

The rule that keeps this safe: client-facing Artifacts get read-only credentials. Write access to your systems always requires explicit confirmation.

How should I price consulting when AI does the delivery?

The consultant billing $200 an hour who compresses a 40-hour audit to 4 hours just converted AI into a pay cut. The resolution is to price the audit at its decision value, $7,000 to $8,000, and let delivery time become your margin.

The productized architecture runs in 3 tiers. Tier 1 is fixed-scope intelligence products: a competitive brief for $1,500 to $3,000 in 48 hours, and a market entry analysis for $3,000 to $6,000 in 5 days. Tier 2 is the retained intelligence service at $3,000 to $8,000 per month, built on the monitoring and synthesis plays, and requires 8 to 15 hours of your time. Tier 3 is the outcome-defined transformation engagement, priced at $25,000 to $75,000, based on value delivered rather than the $5,000 to $15,000 the same work would have earned under hourly rates.

The institutional signal points the same direction. Sequoia’s services-as-software thesis notes $6 flows to services for every $1 to software, and argues the winners will sell completed work, not tools. DEMG’s analysis documents outcome-based practices commanding 5 to 6x EBITDA at exit against 4 to 4.5x for hourly firms. Digital Applied’s benchmarks show solo practitioners reaching $200 to $500 effective hourly rates on productized offerings. And there is an ethical floor here: if you bill time-and-materials, you bill the 4 hours, not the 40. Outcome pricing eliminates the ambiguity.

What are the risks of using AI in client work?

Hallucination: The HBS jagged frontier experiment found consultants using AI outside its competence were 19 percentage points more likely to produce wrong answers. The Stanford RegLab study of legal AI tools found error rates of 17 to 33 percent in tools marketed as hallucination-free. Barclay Damon’s legal analysis states the standard: the legal risk lies not in the existence of hallucinations but in the failure to govern and verify them. Trace every market sizing figure to a primary source. No exceptions.

Confidentiality: Use Claude Team or Enterprise and Perplexity Enterprise Pro for client work. Consumer tiers carry weaker data commitments. For defense clients handling ITAR or CUI material, verify contractual data handling with counsel before uploading anything.

Trust mechanics: MIT Sloan Management Review’s global survey found users who can interpret AI outputs are 2.8x as likely to trust the technology. Your clients follow the same curve. Show sources, show methods, and the deliverable lands.

Workflow fragility: Model updates break Skill behavior, long Project histories dilute context, OAuth tokens expire. Pin models for production workflows, compress old conversations into summary documents, and check MCP connections weekly.

What do the first 90 days look like?

Days 1 to 30, foundation. Build the Firm Voice Project, configure 1 Perplexity Space per active engagement, connect Attio MCP, run your first 5 Deep Research queries. Success: Claude produces on-voice first drafts without style corrections.

Days 31 to 60, automation. Build the Proposal Skill and 2 framework Skills, configure 2 to 3 scheduled monitoring Tasks, start the content engine. Success: proposal drafts in under 10 minutes, intelligence briefs arriving automatically.

Days 61 to 90, repricing. Launch the first productized offering, raise rates on new engagements, ship the first client-facing Artifact. Success: 1 engagement closed at outcome pricing and a measurably higher effective hourly rate.

The order matters. Foundation before automation, automation before repricing. A consultant who reprices before the delivery system exists has sold a promise with no machine behind it.

- j -

FAQ

Is this realistic for a generalist, or only for niche consultants? The stack works for both. The economics favor the niche. Scheduled monitoring, vertically tuned Spaces, and a newsletter thesis all compound faster inside a defined boundary. Weak boundary: business strategy. Strong boundary: GTM strategy for defense tech firms navigating CMMC compliance.

What does the full stack cost? Claude Pro or Team runs $20 to $30 per seat per month; Perplexity Pro $20, with enterprise tiers higher. Attio and a meeting notetaker add $50 to $100. Total tooling sits under $200 a month for a practice billing $15,000 to $50,000 a month. The constraint is build time, not cash.

Do clients care that AI produced the work? Clients pay for judgment and outcomes. You own every deliverable, and your professional accountability is undiminished by who drafted it. The MIT Sloan trust data says the practical move is transparency about the method paired with visible source citations.

Which play should I build first? The Firm Voice Project. Every other play inherits its quality. A proposal Skill without encoded voice produces templates. With it, the output sounds like you on the first pass.

Can I run this on consumer-tier subscriptions? For internal work, yes. For client work, no. Enterprise tiers include the contractual data protections, SOC 2 compliance, and no-training commitments required by client confidentiality.

Does outcome pricing work for existing clients? Reprice at the next natural boundary, a renewal, a new SOW, a scope change. Grandfather current agreements. New engagements start on the new model from day 1.

What breaks first? Verification discipline. The system produces confident, polished, occasionally wrong output at volume. The consultant who stops spot-checking citations ships the error that costs the relationship. Build verification into the workflow, not the calendar.

Become A Paid Subcriber Today To Access All The Assets Below.

Become a Paid Subscrber

Why Enterprise AI Pilots Stall, and the 51 That Didn’t

John Brewton — Sun, 07 Jun 2026 13:13:03 GMT

The operator’s path is rarely a straight line. It crosses open ground, stalls at a failure point, then recovers to deliver a tangible return on investment, hopefully. The route from deployed to value is an organizational journey, and the 51 companies examined in the Stanford study, analyzed in this piece, reached the end by managing the organizational and human path, not by buying a better model.

Too Long → Didn’t Read

Most enterprise AI pilots stall for organizational reasons, not technical ones. A 2025 MIT NANDA study found 95% of generative AI pilots produce no measurable financial impact. The Stanford Digital Economy Lab went the other way and documented 51 successful deployments. The model was never the differentiator. The organization was. This piece walks through where the real work sits and where to spend the next dollar of AI budget. It is about a six-minute read.

What you will find below:

Why do most enterprise AI pilots fail?
- The 95% versus the 51, and why the model is not the variable.
What is the hardest part of deploying enterprise AI?
- 77% of the hardest work is invisible, and 61% of winners failed first.
Why do similar AI projects take weeks at one company and years at another?
- The accelerators and the drags, none of them technical.
What does effective executive sponsorship of AI involve?
- Clearing blockers, permitting failure, and managing staff-function resistance.
Does experience with AI compound over time?
- The data says yes, and the gap is widening.
Does enterprise AI reduce headcount, and where should AI budget go?
- Reduction is a choice, not a verdict.
Frequently asked questions and a full appendix with sources and downloads.

Two companion research reports are free to download: Inside 51 Successful AI Deployments and The Learning Curve.

Are you new to Operating?

You might enjoy these recent reader favorites:

Subscribe now

Why do most enterprise AI pilots fail?

Most coverage of enterprise AI starts with the failures. MIT’s NANDA initiative found that 95% of generative AI pilots produce no measurable financial impact. Stanford’s Digital Economy Lab ran the experiment in reverse. Elisa Pereira, Alvin Wang Graylin, and Erik Brynjolfsson documented 51 deployments that delivered real value, drawn from 41 organizations across 9 industries, 7 countries, and 5 regions, covering more than one million workers. Then they asked what those organizations did differently (Figure 2).

Figure 2 · The failure rate and the exceptions

Two numbers set the stakes. 95% is the failure rate from the MIT study. 51 is the count of deployments Stanford chose to study precisely because they worked.

The answer aligns across all 11 chapters of their report. The model was never the differentiator. The organization was. The technology is the easy, increasingly commoditized part. The hard, durable, value-creating work is organizational. It is human. What great news for all us humans. The bottleneck is the org, not the technology. Both studies, from opposite directions, reach the same conclusion: AI fails at the org level, not the model.

What is the hardest part of deploying enterprise AI?

When practitioners were asked to name the hardest challenges, they did not mention technology. They named change management, data quality, and process redesign. Those intangible costs accounted for 77% of the hardest work. Teams described the technology itself as the easiest piece of the build (Figure 3).

Figure 3 · The invisible 77%

The visible 23% is the technology, the part a business case usually budgets for. The 77% below the line is change management, data quality, and process redesign, the work that decides the outcome and rarely makes the spreadsheet. The yellow rule marks the waterline between what leaders see and what actually costs them.

The cost of the business case missed is the failure that came first. 61% of successful projects had a prior failed AI project. That sunk cost never appears in the winning project’s ROI, yet it was often essential to the win. First attempts failed for consistent reasons. Teams pointed AI at broken workflows. Technical groups were led without business ownership. Leaders assumed the model would fix problems that required redesigning the work.

The data backs a counterintuitive point about readiness. Only 6% of deployments had data fully ready for AI, yet 91% processed unstructured data and 88% made previously inaccessible data usable (Figure 4).

Figure 4 · The data-readiness paradox

The “perfect the data first” instinct is the one this panel dismantles. Only 6% of these teams had ready data, and they shipped anyway. The lesson on the right is the durable one: messy proprietary data you own beats clean data anyone can buy. The operator’s move is to budget for the invisible 77% and to treat the first failure as tuition rather than waste.

Why do similar AI projects take weeks at one company and years at another?

The cleanest proof of the thesis is timing. A fintech migrated millions of lines of legacy code in weeks. A major bank attempting a comparable customer support build reports timelines measured in years. Same technology. Same category of use case. Different organizations, and an outcome gap of more than an order of magnitude (Figure 5).

Figure 5 · Same use case, weeks or years

Two tracks, one technology. The top path is short and direct because the organization was ready. The bottom path winds across the whole frame, detours through a failed pilot, and arrives years later. The visual makes the point a table cannot: the distance between the two is organizational readiness, nothing else.

Three factors consistently accelerated time to value: executive sponsorship in 43% of cases, building on an existing foundation in 32%, and genuine end-user demand in 25%. Four factors consistently slowed projects down, each in roughly a fifth of cases: the learning curve and iteration, data quality and preparation, regulatory and compliance work, and gaps in process documentation. None of those four is a model problem. All four are organizational (Figure 6).

Figure 6 · The forces on time to value

A tug-of-war on a single axis. The yellow vectors pull a project toward ROI, the counter-vectors pull it back toward the stall. Read the labels and the editorial point is unmistakable: every force on both sides is a choice about people and process, not a property of the model.

One pattern was absolute. Every successful project with an identifiable methodology used iteration. None used waterfall. The shape of the work was always start small, learn, expand.

What does effective executive sponsorship of AI involve?

The report draws a sharp line between sponsors who drive results and sponsors who approve budgets. On a four-point engagement scale, only 29% of cases reached Strategic Integration, the level at which AI adoption became a corporate OKR tied to incentives. The seven cases that achieved organization-wide change all reached that level. None reached it by writing a check (Figure 7).

Figure 7 · The sponsorship ascent

Sponsorship is drawn as a climb because that is what it is. Most leaders stop at Active Steering, the 58% rung. The summit, Strategic Integration, is where AI becomes a corporate objective tied to incentives, and it is the only rung where organization-wide change happened. The red mark shows where most stall out.

Effective sponsors do specific things. They clear blockers weekly. They bridge the business and technical teams. They give teams permission to fail, and they mean it. In no case studied was anyone punished for a failed AI initiative. 73% of teams started small on purpose, and 63% framed their pilots as experiments rather than launches.

Resistance came from a place most leaders do not expect. Staff functions, not frontline workers, were the most frequent source. Legal, HR, Risk, and Compliance accounted for 35% of resistance. Internal end-users accounted for 23%. Frontline fear of replacement appeared in only two cases. Staff functions have the authority to slow a project regardless of executive support, and they respond to governance rather than persuasion. Bring them in early with a mandate, and several of them will convert from blockers to enablers.

The same discipline shows up in the amount of human oversight each function needs. Escalation models, where AI handles the work and humans review exceptions, delivered the highest gains. Approval models fit regulated, high-stakes work. The right level depends on the task, not the technology (Figure 8).

Figure 8 · Oversight by function

One graphic carries six data points and a design principle. Functions sit in the oversight band that fits them, plotted by the productivity gain they delivered. The takeaway for an operator is to stop asking how much to trust AI in general and start matching the level of oversight to the task at hand.

Does experience with AI compound over time?

The Stanford data describes organizations. A second dataset describes individuals, and it points in the same direction. Anthropic’s Economic Index studied one million conversations and found that experience with AI is self-reinforcing (Figure 9).

Figure 9 · The advantage compounds

The curve climbs with tenure. Newcomers succeed about two-thirds of the time, seasoned users closer to three-quarters, and roughly four points of that gap survive every statistical control. The editorial reading is that skill with AI is earned through use, which means the advantage is cumulative and the latecomer cannot simply buy it.

Users with six or more months of tenure succeed more often than newer users. Their task success rate is 73.1%, compared with 66.7% for newer users. That edge survives strict controls. Comparing high-and-low-tenure users within the same narrowly defined task, and then adding controls for model, language, use case, and country, leaves a success gap of about 4 percentage points. The difference is not the task they pick. The difference is the skill they built through use.

Experienced users also calibrate. They send their hardest, most valuable work to the most capable model. For every additional $10 in hourly wage for a task, Opus's share increases by 1.5 points on Claude.ai and by 2.8 points on the API. Programmatic users calibrate roughly twice as hard as web users (Figure 10).

Figure 10 · Calibration by task value

Two slopes, one behavior. As the value of a task rises, sophisticated users reach for the more capable model, and developers working through the API do it about twice as steeply as web users. The slope is a portrait of judgment: knowing which model to spend on is itself a learned skill.

Adoption is uneven and staying that way. The top 20 countries account for 48% of per-capita usage, up from 45%, while usage within the United States slowly converges. Read the two datasets together, and the conclusion is plain. Ready organizations deploy in weeks. Experienced users succeed more often. The gap between early movers and everyone else widens with time. The cost of waiting is not flat (Figure 11).

Figure 11 · Adoption concentration ·

Two arrows moving in opposite directions. Across countries the heaviest users pull further ahead, while inside the United States usage slowly evens out. The composition holds both truths at once and underlines the theme: the benefits of early adoption are concentrating, not spreading evenly.

Subscribe now

Does enterprise AI reduce headcount, and where should AI budget go?

Step back and the throughline resolves. For 42% of implementations, model choice was fully interchangeable. The durable advantage sat in the orchestration layer and the organization around it, not in the foundation model. As open-source models close the performance gap, the differentiator shifts from which model you run to what data you feed it and how well your organization works. Model choice only starts to matter as task complexity rises (Figure 12).

Figure 12 · The commodity boundary

The boundary line rises with complexity. For routine work the model is interchangeable, so the smart money is spent elsewhere. For advanced, consequential work the model becomes a real differentiator. The chart tells an operator exactly where model choice stops being a commodity and starts being a decision.

That reframes the question most leaders are actually asking: what AI does to headcount. Reduction was the largest single outcome at 45%. It was also a minority. The other 55% of deployments redeployed people to higher-value work, avoided planned hiring, or reinvested the gains into the roadmap. Reduction is a choice an organization makes, not a verdict the technology delivers. The forward-looking caveat is real and worth stating plainly. ADP payroll data show that early-career workers aged 22 to 25 in AI-exposed occupations have already experienced a 16% relative decline in employment since late 2022. The 45% figure may be a floor (Figure 13).

Figure 13 · The headcount fork

One gain, two destinations. The same productivity improvement forks into the 55% that reinvest or redeploy and the 45% that cut. Drawing it as a fork rather than a fait accompli is the point: the technology produces the gain, the organization chooses what to do with it.

One more pattern deserves attention because it is where the next advantage sits. Agentic AI appeared in only 20% of deployments but delivered a 71% median productivity gain, the highest of any approach. The least-used approach pays the most (Figure 14).

Figure 14 · The agentic quadrant

The quadrant inverts the usual intuition. Agentic AI is the least adopted approach and the highest paying, sitting alone in the top-left while the safer, more common approaches cluster lower and to the right. For a leader looking for the next edge, the empty space around the agentic node is the opportunity.

Subscribe now

The companies that won pointed AI at revenue and new capability, not only at cost. They treated the model as the commodity it is becoming and devoted their utmost effort to the organization. The next dollar of AI budget should go to the same place, because that is where the bottleneck has been the whole time (Figure 15).

Figure 15 · The numbers that matter

A single board to carry away. Six numbers, each one a section of the argument compressed to its figure. It works as the closing slide and as the social card: if a reader keeps only one image, this is the one that holds the whole case.

Are you new to Operating?

You might enjoy these recent reader favorites:

Subscribe now

Frequently asked questions

Why do most enterprise AI pilots fail? They fail for organizational reasons, not technical ones. The MIT NANDA study reported a 95% failure rate and attributed it to workflow integration and misaligned incentives. Stanford’s 51 successes confirm the inverse: the model was never the differentiator.

Is the AI model the most important factor in enterprise AI success? No. For 42% of implementations, the model was fully interchangeable. The durable advantage sits in the orchestration layer, the proprietary data, and the surrounding organization. Model choice only becomes decisive as task complexity rises.

How long does enterprise AI take to deliver ROI? It ranges from weeks to years for the same use case. The variable is organizational: executive sponsorship, an existing foundation, and genuine user demand accelerate it. Learning curve, data prep, compliance, and documentation gaps slow it down.

Who resists enterprise AI adoption the most? Staff functions, not frontline workers. Legal, HR, Risk, and Compliance accounted for 35% of resistance, ahead of end-users at 23%. They respond to a governance role and a clear mandate, not persuasion.

Does enterprise AI cause layoffs? Sometimes. Reduction was the largest single outcome at 45%, but most deployments redeployed people, avoided hiring, or reinvested the gains. Reduction is a choice, not an inevitability, though early-career roles are already declining 16% in AI-exposed occupations.

Does agentic AI deliver better results? Yes, where it fits. Agentic implementations delivered a 71% median productivity gain, the highest of any approach, but appeared in only 20% of cases. It suits high-volume, repetitive work with clear success criteria and recoverable errors.

How clean does enterprise data need to be for AI? Cleaner is better, but perfection is not the gate. Only 6% of deployments had fully ready data, yet 91% processed unstructured data and 88% made inaccessible data usable. Access matters more than centralization, and proprietary data is the durable moat.

Does skill with AI compound over time? Yes. High-tenure users succeed 73.1% of the time, compared with 66.7% for newcomers, and the edge survives strict controls. The advantage accrues to those who start early and build the habit.

Appendix

Primary sources

Pereira, E., Graylin, A. W., and Brynjolfsson, E. (2026). The Enterprise AI Playbook: Lessons from 51 Successful Deployments. Stanford Digital Economy Lab. Publication page · Full PDF
Massenkoff, M., Lyubich, E., McCrory, P., Appel, R., and Heller, R. (2026). The Anthropic Economic Index Report: Learning Curves. Anthropic. Report
MIT Project NANDA (2025). The GenAI Divide: State of AI in Business 2025. Reported figure: 95% of generative AI pilots produce no measurable financial impact. Coverage via Fortune & MIT Media Lab

Supporting references

Brynjolfsson, E., Rock, D., and Syverson, C. (2021). The Productivity J-Curve. National Bureau of Economic Research. The framework behind the report’s macro context.
ADP payroll data, cited in the Stanford report: early-career workers aged 22 to 25 in AI-exposed occupations saw a 16% relative employment decline since late 2022.

Books by the authors

Erik Brynjolfsson and Andrew McAfee. The Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Technologies. W. W. Norton, 2014. Buy on Amazon
Andrew McAfee and Erik Brynjolfsson. Machine, Platform, Crowd: Harnessing Our Digital Future. W. W. Norton, 2017. Buy on Amazon
Alvin Wang Graylin and Louis Rosenberg. Our Next Reality: How the AI-powered Metaverse Will Reshape the World. Foreword by Neal Stephenson. Nicholas Brealey, 2024. Buy on Amazon

Download the companion research reports

Inside 51 Successful AI Deployments. The Stanford playbook distilled: 11 headline findings, all 34 detailed findings by chapter, and a 10-point operator checklist. (Report-1_Inside-51-Successful-AI-Deployments.pdf)
The Learning Curve. The Anthropic Economic Index data on adoption, model selection, and why experience compounds. (Report-2_The-Learning-Curve.pdf)

John Brewton · Founder · Operating · 6AEP

About the author

The 17 Questions That Tell You Whether You Are AI Fluent

John Brewton — Tue, 02 Jun 2026 11:37:11 GMT

TL · DR (The 30-second version)

AI fluency is not knowing what AI can do. It is changing how your work actually gets done. Knowing the capability exists changes nothing on its own.
The research splits in two. Stanford finds 14 to 34 percent task-level gains. NBER finds 89 percent of executives see no enterprise boost. McKinsey finds only 6 percent capture real EBIT, and the one behavior that separates them is workflow redesign. The space between the two is the operator’s opportunity.
This is a 17-question field test, grouped under seven characteristics of the fluent operator. Each question is a binary, and the standard for yes is that you could show the artifact, the schedule, or the screen recording in the next ten minutes.
Score yourself 0 to 17. Four tiers: Curious, Practicing, Operating, Fluent. Most operators land at four or five and assume they are close. They are not.
The worksheet is at the end. Run it on yourself, then on the person who reports to you. The pattern of where you both stall is your build order.

If you have 30 seconds, read this. If you have 12 minutes, read the rest.

Subscribe now

There is a quiet split within companies, and research has now measured both sides of the schism.

At the level of a single task, the gains are real and large. A Stanford study of 5,179 customer support agents found that access to a generative AI assistant raised output 14 percent on average and 34 percent for the least experienced workers. At the level of the whole enterprise, the same technology shows almost nothing. A National Bureau of Economic Research survey of 6,000 executives found 89 percent reporting no productivity gain from AI over three years. Goldman Sachs Research reached the same conclusion in 2026: no meaningful relationship between AI and productivity at the economy-wide level, alongside a median 30 percent gain in two narrow places, software engineering and customer service. The task data and the enterprise data are not the same data. The space between them is the operator’s opportunity.

McKinsey put a number on who closes that space. In its late 2025 State of AI, 88 percent of organizations report using AI, and only 6 percent qualify as high performers capturing 5 percent or more of EBIT from it. The single strongest correlation with that impact is not model choice or budget. It is fundamental workflow redesign. High performers are nearly three times as likely as everyone else to say they have rebuilt how the work is done.

That finding is the whole argument of this piece. AI fluency is not knowing what AI can do. It is changing how your work actually gets done. Knowing the capability exists produces nothing. MIT found that 95 percent of enterprise AI pilots returned no measurable result, and the cause was not the model. It was the failure to redesign the work around it. Fluency lives in the work itself, not in the vocabulary around it.

So I built a test for it. Seventeen questions, grouped under seven characteristics of the fluent operator. Each one is grounded in the institutional research record, and each one has a single hard standard.

How to take this

Each question is a binary. Yes or no. No partials and no “kind of.”

The standard for a yes is strict. You could show the artifact, the schedule, or the screen recording in the next ten minutes. If your answer is a description of what is possible, the answer is no. If your answer is a piece of work that already runs differently, the answer is yes.

Count your yes answers as you go. There are four tiers at the bottom. Most operators assume they are most of the way there and land lower than they expect. That gap is the point of the exercise.

Subscribe now

Section I · Do The Work

The job runs differently because AI is inside it.

1. Name one recurring deliverable you own that you now produce through an AI-assisted workflow you did not use twelve months ago. Could you show the before-and-after side by side?

A yes is a specific deliverable, a brief, a deck, a model, a memo, that is genuinely produced differently now. This is the load-bearing question. The Stanford gains showed up only for workers who changed how the task was done, not for workers who knew the tool existed.

2. Has your role, your weekly cadence, or your standard output changed in the last 90 days because AI is now inside the work, not adjacent to it?

A yes means something structural shifted, not that a few steps got faster. This is where the McKinsey finding bites hardest. Workflow redesign is the behavior that separates the 6 percent who capture real value from everyone else. The MIT pilots that failed bolted AI onto the old process. A faster version of a broken step is still a broken step.

Section II · Wired into the stack

No separate environment to switch into.

3. Is your AI connected to the apps you already live in, email, calendar, files, CRM, so you never copy and paste between a chat window and your real work?

A yes means the model reaches your systems of record directly. Andreessen Horowitz’s 2025 survey of 100 enterprise CIOs found enterprises expecting roughly 75 percent growth in model spend over the next year, and one CIO put it plainly: what I spent in 2023 I now spend in a week. The money is moving toward systems that act inside connected tools, not systems that chat.

4. Have you installed the six core connectors, email, calendar, CRM, Notion or Drive, Slack, and accounting, and can you name the last action each one took on your behalf?

A yes is four or more connectors live and in daily use. This is where the NBER number bites. Most of those 6,000 executives adopted tools and never connected them to the real work, and the productivity gain never arrived. Adoption without integration produces nothing.

Section III · Communications Hub

Comms and calendar run themselves.

5. Does a daily or weekly brief arrive on your screen without you asking for it, covering your inbox, your calendar, conflicts, and the highest-impact items of the day?

A yes is something useful that shows up unprompted. Goldman Sachs Research finds AI’s measured gains run near 30 percent but stay confined to narrow, localized use cases. The unprompted morning brief is exactly that kind of case. It is small, repeated, and it compounds.

6. Is your inbox triaged and your replies drafted before you open it most mornings, with scheduling that resolves without your direct intervention?

A yes means the rhythm of your communication runs with light touch. Thomson Reuters surveyed 2,275 professionals and found they expect AI to save them 5 hours a week, about 240 hours a year, worth roughly 19,000 dollars per person. Inbox triage is that routine work, and it is the first place most operators feel the week get lighter.

Section IV · Analysis

Analysis collapses to minutes.

7. Have you produced a board-ready or client-ready brief inside a single hour in the last 30 days, using a curated research workflow rather than an open chat?

A yes is a real artifact produced fast and clean. Stanford’s AI Index puts organizational adoption at 78 percent in 2024, with the clearest output gains in drafting and data preparation. A disciplined research brief is that gain, productionized into your week.

8. Have you constrained your research to a defined universe of Tier 1 publications, Tier 2 academic centers, and Tier 3 institutional shops, and can you name three sources you have excluded?

A yes means you build the research environment before you ask the question. Constraining the model to a defined source universe moves the signal-to-noise ratio from roughly 20 percent to roughly 90 percent in one move. The fluent operator never lets the model choose the sources. (This article is itself an example. Every link in it sits inside a curated source universe, and the appendix shows the architecture.)

9. In the last week, did you drop a CSV or pull live data, describe a question in plain English, and iterate to a publication-quality chart in under an hour?

A yes is the chart-building loop running without an analyst in the chain. In the Harvard and BCG study of 758 consultants, work inside AI’s reliable zone came back 40 percent higher in quality and 25 percent faster. The chart loop sits squarely inside that zone.

Section V · Agents

Headcount stops being the only way to add capacity.

10. Do you have at least one scheduled agent, written as a one-page spec with a defined job, defined access, and a defined report-back, that ran last week without you starting it?

A yes is one agent doing real work on a schedule. Harvard Business Review found that 91 percent of executives expect agentic AI to reshape work, and only 13 percent have shipped an agent. The agent count is the clearest marker of where you sit in that gap.

11. Do you have between four and seven agents carrying rhythm work, briefing, pipeline, follow-up, reconciliation, that used to sit on your desk?

A yes is a small team of agents holding the recurring load. Sequoia argues the next trillion-dollar company will not sell software but the work itself, delivered by agents, and draws a sharp line between a copilot that helps you work and an autopilot that returns the finished work. The operator who already runs a few autopilots is early to the thing the capital is chasing.

12. When an agent produces a bad outcome, do you know who owns the action, where the kill switch is, and how to scope its permissions tighter?

A yes means you run agents with discipline. HBR warns against treating an agent as an employee. Run each one with scoped permissions, observability, a kill switch, and a single named owner per action. The trust gap is the reason this matters. Only 6 percent of companies fully trust agents with core processes, and 48 percent name a lack of skills as the top barrier. Discipline is the skill that closes that gap.

Section VI · Standard

Speed and judgment improve together.

13. In the last 30 days, did you catch a confident, polished AI output that was wrong on a fact, a number, or a source, and correct it before it left your desk?

A yes means your judgment is active, not delegated. AI lifts some tasks and quietly degrades others of equal difficulty, with no signal telling you which. In the Harvard and BCG trial, work outside the model’s reliable frontier came back measurably worse. The fluent pattern keeps the human in the loop. Anthropic’s Economic Index found augmentation, where a person iterates with the model as a thinking partner, back above half of all use at 52 percent, ahead of hands-off automation. Your judgment is the filter, and a polished wrong answer is more dangerous than an obviously bad one.

14. Is the faster output you now produce holding up to scrutiny from a CEO, a board, a client, or an investor, with no cost to quality, sourcing, or defensibility?

A yes means speed did not come at the cost of standard. Robert Solow observed in the New York Times in 1987 that the computer age was visible everywhere except the productivity statistics. The lag closed when the output held up, not when it merely got faster. Fluency is faster output that survives scrutiny.

15. Have you cleared desk-work hours against a baseline you actually measured, not estimated, with 12 to 18 per week as the operator benchmark?

A yes requires a real measurement, not a feeling. A controlled trial by METR found that experienced developers were 19 percent slower with AI while believing they were 20 percent faster. Felt time saved is not measured time saved. If you have not measured, you cannot know which side of that gap you are on.

Section VII · Portability and leadership

Platform-agnostic. Teachable downward.

16. Could you rebuild your research workflow, your dashboards, and your agent team on a different platform inside a weekend if the field shifted tomorrow?

A yes means the capability lives in your habits, not in one vendor’s interface. Andreessen Horowitz found 37 percent of enterprises now run five or more models in production, up from 29 percent a year earlier, spreading spend on purpose to avoid lock-in. Fluency built on transferable habits survives the field moving under it.

17. Have you taught at least one other person to install and run one of your three builds on their own work, so fluency is a leadership capability and not a personal hack?

A yes is the capability installed in someone else. Mentions of generative-AI skills in job postings roughly quadrupled in a single year, from about 16,000 to 66,000. Fluency is now a hiring requirement, which makes the ability to teach it a leadership position rather than a personal trick.

Subscribe now

Count Your Yes Answers

0 to 5 · Curious. You know the territory. You do not yet operate inside it. Start with the research architecture this weekend and connect the core six tools.

6 to 10 · Practicing. The builds are underway. The week has not yet been reshaped. The gap is usually the operator skills and the first scheduled agent. This is where most people stall and assume they have finished.

11 to 14 · Operating. The work runs differently. Time is being returned to judgment. Add agents until four to seven carry the rhythm work, then teach one other person.

15 to 17 · Fluent. The capability is portable, teachable, and inside the work. Install it across your team. Fluency is now a leadership position.

Most operators answer yes to four or five and assume they are close. The distance between five and seventeen is the entire competitive position of the job. The cost of staying at five for another quarter is the part nobody is pricing correctly yet.

Take the assessment

I have built these seventeen questions into a one-page worksheet. It is a fillable PDF with a yes and no box for every question, the four scoring tiers, and the build sequence to close your gaps. You can answer it on screen or print it and work through it with your team.

Download the AI Fluency Diagnostic worksheet (PDF)

Run it on yourself first. Then run it on the person who reports to you. The pattern of where you both stall tells you exactly where the next 30 days of work goes.

If you want to do the builds rather than just score them, I run a three-week sprint called Oper(AI)te that walks operators through all three, live, on their own work. Details are at unfazedfounder.com/operaite.

Frequently asked questions

What is AI fluency?

It is the ability to work differently because AI is in the loop. It is not knowledge of what AI can do. A fluent operator produces real deliverables through AI, runs research in minutes, and hands rhythm work to scheduled agents. MIT found 95 percent of enterprise pilots returned nothing measurable, because the work was never redesigned. Fluency is the redesign.

How is this different from being good at prompting?

Prompting is one input. Fluency is whether the work itself runs differently. You can write excellent prompts and still copy and paste between a chat window and your real job. This test measures the output and the system around it, not the prompt.

Do I need to be technical to score well?

No. The questions measure whether your work runs differently, not whether you can code. The operator’s skills are sequencing, source discipline, and judgment. The platform handles the technical layer.

What counts as a yes?

A yes is a piece of work you could show in the next ten minutes: the artifact, the schedule, or the screen recording. A description of what is possible is a no. There are no partials.

I scored lower than I expected. Is that normal?

Yes. Most operators answer yes to four or five and assume they are most of the way there. The distance from five to seventeen is the actual job. The questions you stall on are your build order.

Which gap should I close first?

Start with the research architecture and the core six connectors this weekend. Then stand up one scheduled agent. The four-tier read lists the sequence, and the worksheet repeats it.

Does the AI platform matter?

The capability spans Claude, ChatGPT, Gemini, and Perplexity and should move among them. Question 16 tests exactly that. Enterprises now run multiple models on purpose to avoid lock-in. Fluency that lives in one vendor’s interface is fragile.

How often should I retake it?

Every 90 days. That is the mark where the time savings stabilize, and a quarterly retake shows whether the builds held or drifted.

Is the worksheet free?

Yes. Download it below, run it on yourself, and work through it with your team.

Appendix · The research behind the assessment

Every claim in this article is sourced from a curated universe of Tier 1 publications, Tier 2 academic centers, and Tier 3 institutional research shops. The appendix organizes that record so you can pull the threads yourself.

A · The findings that anchor the assessment

1. Task-level productivity gains run 14 to 34 percent. Brynjolfsson, Li, and Raymond studied 5,179 support agents. Output rose 14 percent on average and 34 percent for the least experienced. [PEER-REVIEWED] Stanford SIEPR · Generative AI at Work

2. Enterprise-level surveys report no productivity boost. A survey of 6,000 executives across four countries found 89 percent reporting no productivity impact over three years. [EDITORIAL] NBER executive survey, via Fortune

3. Value concentrates in the 6 percent who redesign the work. McKinsey found 88 percent of organizations use AI, only 6 percent capture 5 percent or more of EBIT from it, and the strongest correlate of impact is fundamental workflow redesign. [INSTITUTIONAL] McKinsey · The State of AI, 2025

4. Adoption without redesign returns nothing. MIT’s NANDA initiative found 95 percent of enterprise AI pilots delivered no measurable P&L result. The root barrier was learning and workflow fit, not infrastructure or talent. [EDITORIAL] MIT NANDA · The GenAI Divide, via Fortune · full report PDF

5. The skill is in production, and the adoption gap is wide. Stanford’s AI Index puts organizational adoption at 78 percent in 2024, with gen-AI skill mentions in job postings rising from about 16,000 to 66,000 in a single year. [PEER-REVIEWED] Stanford HAI · 2025 AI Index Report

6. The capital is voting on agents. a16z’s survey of 100 enterprise CIOs found model spend growing roughly 75 percent year over year, with 37 percent of enterprises now running five or more models in production. [INSTITUTIONAL] Andreessen Horowitz · 100 Enterprise CIOs, 2025

B · Complicating evidence

Three findings push back on the simple version of the story. They belong in any honest read of the record.

AI can slow experienced people down. A randomized controlled trial found experienced developers were 19 percent slower with AI while believing they were 20 percent faster. [PEER-REVIEWED] METR · randomized controlled trial, July 2025

The frontier is jagged. AI improves some tasks 40 percent and degrades others of equal difficulty, inside the same workflow, with no signal which is which. [PEER-REVIEWED] Dell’Acqua, Mollick et al. · Navigating the Jagged Technological Frontier, Organization Science 2026 · SSRN working paper

Trust is the live constraint on agents. Only 6 percent of companies fully trust agents with core processes, 43 percent restrict them to routine work, and 48 percent cite a lack of skills as the top barrier. Run agents with scoped permissions, observability, kill switches, and a named owner. [EDITORIAL] HBR research, via Fortune · Why You Shouldn’t Treat AI Agents Like Employees, HBR 2026

C · The newest signals (2026)

Goldman sharpened the gap. Its 2026 update found no economy-wide productivity relationship, with the 30 percent gains localized to software engineering and customer service, even as AI capex climbed to a forecast 667 billion dollars. [INSTITUTIONAL] Goldman Sachs Research, via Fortune, March 2026 · Goldman · 7% global GDP forecast

Augmentation pulled back ahead of automation. Anthropic’s Economic Index found augmentation, where a person iterates with the model as a thinking partner, back above half of all use at 52 percent, ahead of hands-off automation. The judgment stays with the operator. [INSTITUTIONAL] Anthropic Economic Index, January 2026

Copilots are becoming autopilots. Sequoia argues the next trillion-dollar company will sell the finished work, not the tool, and notes that more tasks are now started by agents than by people. [INSTITUTIONAL] Sequoia Capital · Services: The New Software

The time math has a price. Thomson Reuters found professionals expect AI to save 5 hours a week, about 240 hours a year, worth roughly 19,000 dollars per person. [INSTITUTIONAL] Thomson Reuters · Future of Professionals, 2025

D · Historical parallel

The gap between task-level AI studies and enterprise-level surveys looks like the Solow paradox of the late 1980s. Robert Solow, writing in the New York Times in 1987, observed that you can see the computer age everywhere except in the productivity statistics. The lag stretched roughly a decade before the late-1990s acceleration. The operators who built the workflow during that lag captured the gains when the macro numbers finally turned.

E · Sources by tier

The elite-research architecture behind this assessment. Every source above sits in one of these three tiers.

Tier 1 · Elite publications

Harvard Business Review · Why You Shouldn’t Treat AI Agents Like Employees [EDITORIAL]
Harvard Business School Working Knowledge · Leadership in an Agentic AI World [EDITORIAL]
Fortune · NBER executive survey coverage [EDITORIAL]
Fortune · MIT GenAI Divide coverage [EDITORIAL]
Fortune · Goldman 2026 productivity coverage [EDITORIAL]
Fortune · HBR agent-trust coverage [EDITORIAL]

Tier 2 · Academic research centers

Stanford SIEPR · Generative AI at Work [PEER-REVIEWED]
Stanford HAI · 2025 AI Index Report [PEER-REVIEWED]
Harvard and BCG · Navigating the Jagged Technological Frontier, Organization Science 2026 [PEER-REVIEWED]
MIT NANDA · The GenAI Divide, State of AI in Business 2025 [PEER-REVIEWED]
METR · Early-2025 AI and Developer Productivity [PEER-REVIEWED]

Tier 3 · Institutional research

McKinsey · The State of AI, 2025 [INSTITUTIONAL]
Goldman Sachs Research · AI and GDP [INSTITUTIONAL]
Andreessen Horowitz · 100 Enterprise CIOs, 2025 [INSTITUTIONAL]
Sequoia Capital · Services: The New Software [INSTITUTIONAL]
Anthropic · Economic Index, January 2026 [INSTITUTIONAL]
Thomson Reuters · Future of Professionals, 2025 [INSTITUTIONAL]

The source architecture above is the elite-research skill described in Question 8, applied to its own subject. Curate the universe, name the sources, exclude the rest.

John Brewton · Founder · Operating · 6AEP · Washington, DC

Build Your Team of Agents

John Brewton — Fri, 29 May 2026 11:46:02 GMT

Subscribe now

Build Your Own Analyst With Claude

John Brewton — Fri, 29 May 2026 11:43:37 GMT

Subscribe now

Build Your Claude Research Lab

John Brewton — Fri, 29 May 2026 11:40:21 GMT

Subscribe now

Three Builds To Become An AI Fluent Operator

John Brewton — Thu, 28 May 2026 11:17:52 GMT

Subscribe now

TL · DR (The 30-second version)

AI fluency is not a skill. It is three builds an operator installs once and runs on for the rest of their career.

Build One: The Research Architecture - Constrain Claude to a curated universe of Tier 1 publications, Tier 2 academic centers, and Tier 3 institutional research shops. Name the mode and output format before the question. Signal-to-noise ratio climbs from 20 percent to 90 percent in one move.
Build Two: Charts and Dashboards - Drop a CSV, describe the question in plain English, iterate in 20-second cycles. The build-dashboard skill produces a self-contained HTML report. The analyst bottleneck collapses.
Build Three: The Team of Agents - Install six core connectors. Write each agent as a one-page spec. Set the schedule. Direct, refine, add. Four to seven agents inside 30 days carry the rhythm work.
The math - A senior operator carries roughly 50 hours of weekly admin load. The three builds compound to clear 12 to 18 hours inside 90 days.
The window - Stanford finds 14 to 34 percent task-level productivity gains. NBER finds 89 percent of executives reporting zero enterprise productivity boost. HBR finds 91 percent of executives believe agentic AI will reshape work but only 13 percent have deployed agents. The space between is the operator opportunity.

If you only have 30 seconds, read this. If you have 20 minutes, read the rest.

Subscribe now

What is AI fluency for a senior operator?

There is a quiet thing happening inside companies right now, and the operators who are paying attention have already named it.

Something that took three days last year takes an afternoon now. A junior analyst with the right setup produces work that a senior used to own. The board deck that ate two evenings gets drafted before lunch.

Nobody is saying this out loud in the all-hands.

Everybody is noticing.

The research backs the pattern. A Stanford Institute for Economic Policy Research working paper of 5,179 customer support agents found access to a generative AI assistant raised productivity by 14 percent on average and by 34 percent for the lowest-skilled quartile. The 2025 Stanford AI Index reports the same pattern across drafting, customer communication, data preparation, and software development. The lift is real, and it concentrates where the work has been the heaviest.

The strange part is what the same research finds at the enterprise level. A National Bureau of Economic Research survey of 6,000 executives across the US, UK, Germany, and Australia reported 89 percent of them are seeing no productivity boost from AI inside their companies. The task-level data and the enterprise-level data are not the same data. The gap between them is the operator’s opportunity.

FIGURE 1 · The Productivity Gap

Source: Stanford SIEPR · Generative AI at Work (Brynjolfsson, Li, Raymond). NBER executive survey via Fortune. Chart by John Brewton.

The word for that opportunity is fluency. AI fluency is not knowing what AI can do. It is changing how your work actually gets done.

I am going to walk you through the three builds that take an operator from curious to fluent. I run all three of them within my own companies and within client engagements at 6AEP. Each solves a different administrative problem. Together, they replace about 30 hours of weekly desk work for the average senior operator.

You can build the first one this weekend.

You can build all three in three weeks.

How do you build an AI research workflow that pulls from elite sources?

Most people use AI for research the wrong way. They open a chat, type a question, and accept whatever comes back. The model pulls from wherever, the citations are inconsistent, and the output reads like a polished blog post written by a sophomore.

The fluent move is different. You build the research environment first. Then you ask the question.

FIGURE 2 · The Research Architecture

The architecture I use is called the elite-research skill. It lives in my Claude account as a reusable instruction set that is automatically invoked whenever I ask a research question. The structure has three layers.

Layer one is for source curation. The skill only searches inside a defined universe of top-tier publications, academic centers, and institutional research shops. The publication list includes the New York Times, Financial Times, Wall Street Journal, The Economist, Harvard Business Review, MIT Technology Review, and MIT Sloan Management Review. The academic list includes Harvard, Stanford, Chicago Booth, Princeton, Yale, and LSE. The institutional list includes Goldman Sachs research, Bridgewater Daily Observations, Blackstone insights, AQR, Andreessen Horowitz, and Sequoia Capital. The skill ignores everything else. No Medium posts, no LinkedIn essays, no SEO content marketing dressed up as analysis.

That single constraint changes the output more than anything else. The signal-to-noise ratio goes from roughly 20 percent to roughly 90 percent in one move.

Layer two is mode selection. The skill has four research modes, and the right one depends on what I am actually trying to do. Thesis Research is for stress-testing a position I already hold. Market Intelligence is for understanding a company or sector cold. Trend Scanning is for catching what is moving across an industry over 30 to 90 days. Deep Dive is for the questions where I need everything the institutional record contains on a narrow topic. I name the mode at the top of the prompt and the skill structures the search accordingly.

Layer three is the output format. The skill outputs in one of three shapes. A Research Brief is a 600 to 900-word memo with sourced bullets. An Advisory Intelligence Memo is the longer format I send to clients, around 2,000 words with a recommended action set. A Signal Pack is the lightest format, used for surfacing five to seven data points fast.

The build itself is straightforward. You write the skill as a markdown file with three sections: source list, mode definitions, and output templates. You install it in Claude as a custom skill. The model then automatically invokes the entire architecture every time you ask a research question.

The first time you run a real question through it, the difference is unmistakable. Yesterday I asked the skill to assess a sector trend across the last two years. The output came back in 11 minutes, with 14 sourced citations, all from inside the curated universe. The same question in an unstructured chat would have produced a 400-word summary with three of the citations made up.

That is the entire point of build one. You stop relying on the model to choose your sources and start treating them as part of your operating discipline.

Companion playbook: The Research Architecture · 8-page field guide (PDF). Curate, Mode, Format, Deploy. The full architecture with worked examples and a maintenance cadence.

How do you build charts and dashboards with AI?

The second build solves the analytics problem. Every operator I work with carries some version of the same complaint. They have data in five places, the BI tools their company pays for are too slow to iterate against, and the analyst who could run the query is booked through Friday.

Here is what fluent looks like. You drop a CSV into the chat. You describe the chart you want. Ninety seconds later you have a publication-quality visualization, with formatting you control, that you can drop into a board deck or a strategy memo.

Subscribe now

FIGURE 3 · Charts You Build Yourself

The Goldman Sachs research team has been blunt about where AI gains actually show up at the macro level. In a 2025 update to its productivity tracker, Goldman found no meaningful relationship between AI adoption and productivity at the economy-wide level, with a median 30 percent productivity gain confined to narrow, localized use cases. The chart-building loop is exactly the kind of localized use case the research describes. The lift does not show up in the firm’s headline numbers. It shows up in the work that one operator does on one Friday afternoon.

I want to show you exactly how this works, because the gap between people who know this is possible and people who actually do it is enormous.

Step one. You collect the data. This part is unchanged. You pull from Stripe or QuickBooks or your warehouse, and you save the result as a CSV. If the data is already inside a connected system, you can skip this step entirely and ask the model to fetch it through the connector.

Step two. You describe the chart in plain English. Not the type, the question. Something like: “Show me monthly revenue from 2024 to today, broken out by product line, with a 3-month moving average overlaid.” The model picks the chart type, builds the matplotlib or Plotly code, runs the code in its sandbox, and returns the chart as an image you can save.

Step three. You iterate. This is where the workflow earns its keep. You say “make the moving average dashed, drop the 2024 quarters, and change the title to fit a board format.” The new chart appears in 20 seconds. You repeat that loop until the chart says what you want it to say.

For anything bigger than a single chart, the build is a dashboard. There is a Claude skill called build-dashboard that constructs a self-contained HTML file with KPI cards, filterable charts, and sortable tables. You give it the data and the questions you want answered. It returns a single HTML file you can open in any browser, share with a colleague, or post on your internal wiki. The dashboards are static once built, which makes them easy to version and easy to trust.

The fluent operator runs through this loop two or three times a week. A weekly revenue review takes 14 minutes instead of two hours. A board prep cycle that used to involve four people coordinating across three tools collapses into a single afternoon at the desk.

The skill ceiling on this build is high. The space between you and having this part complete is one weekend.

Companion playbook: Charts and Dashboards · 8-page field guide (PDF). Collect, Describe, Iterate, Dashboard. The data-handoff discipline, the prompt shape, and the iteration loop.

How do you deploy AI agents that work on your behalf?

The third build is the one that changes the shape of the workweek. It is also the one that takes the most setup, which is why most operators stop at the first two and never get here.

An agent is a specialized AI assistant configured once, with a defined job, that runs on request. You direct it. It does the work. It reports back. The work happens 24 hours a day, 7 days a week, and you do not need to be present.

FIGURE 4 · Your Team of Agents

The institutional capital is voting on this build hard. Andreessen Horowitz’s 2025 survey of 100 enterprise CIOs found that average enterprise spend on large language models rose from $4.5 million to $7 million in 24 months, with another 65 percent jump expected to roughly $11.6 million the following year. The fastest-growing developer behavior in the same data is what a16z calls agentic inference, where the model acts across extended sequences rather than single prompts. Sequoia Capital went further. The firm’s 2026 essay Services: The New Software argues the next trillion-dollar company will not sell software tools but the actual work itself, delivered as a service powered by agents.

The adoption story is the inverse of the capital story. A 2026 Harvard Business Review survey reports 91 percent of executives believe agentic AI will reshape work, and 83 percent say effective adoption is essential to competitive position. Only 38 percent feel their organization is prepared. Just 13 percent have integrated agents into real workflows. 56 percent are still piloting under heavy human supervision. The operators building the agent stack right now are competing within a window that institutional data say is unusually wide.

FIGURE 5 · Belief vs Build

Source: HBR 2026 enterprise survey via HBS Working Knowledge. Chart by John Brewton.

You build agents through connectors. A connector is a software bridge that allows the model to communicate with one of your tools. Gmail, Google Calendar, Notion, Slack, HubSpot, QuickBooks, and Google Drive all have connectors that take about 90 seconds each to install. Once a connector is live, every agent you build can read from and write to that system on your behalf, with the permissions you set.

The plugin packages in Claude are what make the agent practical to build. A plugin is a bundle of connectors, skills, and pre-built templates organized around a function. The Sales plugin includes connectors for your CRM and email, plus pre-built skills for account research, call prep, pipeline review, and outreach drafting. The Operations plugin includes process documentation skills, runbook builders, and risk assessment templates. The Productivity plugin includes a memory system and a shared task file that the agents read and update.

Once the plugins are installed, the agent build looks like this.

You pick one piece of work that has been sitting on your desk too long. For most operators, that work is one of four things. The Monday briefing nobody has time to write. The pipeline review that gets done at 11 p.m. on Sunday. The customer follow-up sequence gets skipped half the time. The expense reconciliation that piles up for two weeks.

You write the agent as a one-page instruction. The instruction has three sections. What the agent does. What it has access to. What it sends back. The first version is short. Roughly 200 words.

You set the agent to run on a schedule. The schedule is the second thing that makes the agent practical. A Monday Briefing agent runs at 6 a.m. every Monday, reads your calendar and your inbox, summarizes what is coming, flags conflicts, and drops the result in your messages before you wake up. A Pipeline agent runs at 5 p.m. every Friday, reads HubSpot, flags deals that did not move, and drafts the follow-up emails.

You direct, refine, and add. Inside a month, a senior operator running this build has between 4 and 7 agents handling defined work. The agents are not magic. They do the things you would do if you had four hours a day you do not have.

The first agent feels like a trick. The fourth one feels like the floor.

Companion playbook: Your Team of Agents · 8-page field guide (PDF). Connect, Write, Schedule, Conduct. Connector install order, the one-page agent spec, four starter agents.

How much weekly time does an AI-fluent operator save?

I built all three of these inside the companies I run because the time math is brutal at the senior operator level. A Head of Operations carries roughly 50 hours of unbilled administrative load each week. Research, analysis, communication, scheduling, follow-ups, status updates, briefing prep, and the work of routing other people’s work.

Building one collapses the research load by about 70 percent.
Building two collapses the analysis load by about 80 percent.
Building three collapses the communication, scheduling, and follow-up load by 50 to 90 percent, depending on how aggressive you get with the agent set.

The fluent operator I describe in client engagements clears 12 to 18 hours of desk work per week within 90 days. That is not a productivity story. That is the entire competitive position of the job.

FIGURE 6 · The Operator Time Math

Chart by John Brewton → Data from 6AEP work

The macro forecast points in the same direction. Goldman Sachs Research projects that generative AI will raise global GDP by roughly $7 trillion and lift productivity growth by 1.5 percentage points over ten years. The historical parallel is the Solow paradox of the late 1980s, when computers were visible everywhere except in the productivity statistics for nearly a decade before the late-1990s acceleration. The current gap between Stanford’s task-level findings and the NBER enterprise survey maps onto that lag with high precision. The operators who close the gap first will be running the work when the macro numbers finally show up.

What does an AI-fluent operator’s week actually look like?

Monday: A deck the board needs by Wednesday. You draft it in an hour, with the research agent feeding sourced bullets and the chart builder generating the four visualizations. You spend the second hour writing the narrative.
Tuesday: Three back-to-back calls. Each one ends with the summary, the decisions, and the action items already written and routed by the meeting agent. You took no notes.
Wednesday: The question that meant an afternoon of reading returns a sourced, current answer before lunch. You make the call by 2 p.m.
Thursday: Your team of agents pulled the numbers, flagged what moved, drafted the update, and built the spreadsheets. You edited. You did not build.
Friday: You closed the week ahead and on time, with every task that was pressing on Monday handled, already moving on to next week.

That week is not aspirational. It is what the fluent operator’s calendar looks like now.

Subscribe now

Can you become AI fluent without taking a course?

You can build a version of each of these on your own. The skill files are documented. The connectors are publicly available. The plugin packages ship with Claude. None of this is proprietary.

What is hard is the sequencing. The order in which you build them, the way you scope the first agent, the prompts that work versus the prompts that look like they should work, and the discipline to keep the source list curated as you go. That is the gap between an operator who tinkers with AI and an operator who runs on it.

I am running a three-week sprint in June called Oper(AI)te that walks operators through these three builds, live, on their own work. Six sessions across three weeks, two a week, 90 minutes each. We connect your stack in week one. We build the research and analytics architecture in week two. We build your agent team in week three. You leave with the connected setup, the research workflow, the dashboard pattern, the agent team, and a 30-day plan to add more as the work evolves.

The sprint begins Tuesday, June 16, 2026. Enrollment closes June 15. Details and the seat link at unfazedfounder.com/operaite.

If the three builds make sense and you want to do them on your own, that is the right answer too.

The cost of waiting another quarter is the part nobody is pricing correctly yet.

Become AI-Fluent This Month

Frequently Asked Questions About Becoming AI Fluent

What is AI fluency?

AI fluency is the ability to work differently because AI is in the loop. It is not knowing what AI can do. It is a change in how work is actually produced. A fluent operator drafts decks in an hour, runs research in minutes, and delegates rhythm work to scheduled agents.

How long does it take to become fluent in AI?

The three builds take three weeks if you sequence them correctly. Build One ships in a weekend. Build two ships in a weekend. Build Three takes the remaining three weeks because the agent team requires the most iteration. The 90-day mark is when the time savings stabilize, and the fluent operator clears 12 to 18 hours of weekly desk work.

Do I need to be technical to build these workflows?

No. The three builds use natural-language instruction inside Claude or another major AI platform. Source curation is a markdown file. Chart-building is plain-English prompts. Agent specs are one-page documents. The technical layer is handled by the platform. The operator's skills are sequencing and prompt discipline.

Which AI platform should I use?

The article uses Claude as the worked example because that is the platform I run on. The same builds work on ChatGPT, Gemini, and Perplexity with minor adjustments to connector availability and skill format. Platform-agnostic fluency is the goal. The Oper(AI)te course covers all four.

What is the elite-research skill?

A custom Claude skill that constrains the model to a defined universe of Tier 1 publications, Tier 2 academic centers, and Tier 3 institutional research shops. It enforces source-tier credibility tags on every finding and outputs in one of three formats: Research Brief, Advisory Intelligence Memo, or Signal Pack. The skill is a Markdown file you install once.

What is the difference between a Claude skill and an agent?

A skill is reusable operating logic that Claude invokes inside a chat or task. An agent is a scheduled or triggered task that runs on its own and reports back. Skills shape how Claude works on a question. Agents do the work without you asking each time. Agents commonly invoke skills as part of their job.

Which agent should I build first?

The Monday Briefing agent. It reads your calendar and your inbox each Monday at 6 a.m., summarizes the week ahead, flags conflicts, and lists the three highest-impact items. It is the easiest spec to write, the highest-frequency value, and the agent that most clearly demonstrates how the rest of the agent stack works.

What if my data lives in tools that do not have a Claude connector?

Two options. The first is CSV export, which works for any system that produces a downloadable report. The second is building a custom MCP connector, which takes about 90 minutes for a competent technical user. Most operators do not need the custom connector path. The core six connectors (Gmail, Calendar, CRM, Notion or Drive, Slack, QuickBooks) cover the work of most senior operators.

How is this different from buying an off-the-shelf AI product?

The three builds compose a general-purpose AI infrastructure around your work. An off-the-shelf product compresses a single function (sales prospecting, transcription, scheduling) into a fixed workflow. The compounding advantage of the three builds is that the same connectors, skills, and agents serve multiple jobs and improve as you refine them. The product replaces a task. The builds replace the way you work.

What does it cost to run these workflows?

Platform subscriptions range from $20 to $200 per month, depending on the AI platform and tier you use. Most senior operators recover the cost of the saved desk time in the first week. The Oper(AI)te course is a one-time $4,500 fee and is structured to make sequencing easier rather than to gate access to the workflows themselves.

Become AI-Fluent This Month

What companion materials accompany this article?

Three resources accompany this piece. Take them in any order.

Field-guide playbooks · 8 pages each · PDF

Playbook 01 · The Research Architecture - Curate the source universe. Pick the mode before the question. Three output shapes. Deploy and maintain the skill.
Playbook 02 · Charts and Dashboards - Get the data out of the tool. Describe the question, not the chart. The iteration loop. The dashboard build.
Playbook 03 · Your Team of Agents - Connectors are the bridge. Write the agent as a one-pager. Set the schedule. Conduct the team.

Elite-research companion brief · Markdown

Research Brief · The AI Fluency Gap - Six findings, complicating evidence, historical parallels, narrative hooks, reading list.

Which institutions have studied AI productivity and adoption?

Every claim in this article is sourced from the elite-research universe. Tier 1 publications, Tier 2 academic centers, Tier 3 institutional research shops. The appendix below organizes the source material so you can pull threads yourself.

A · Six findings that anchor the article

1. Task-level productivity gains run 14 to 34 percent. Brynjolfsson, Li, and Raymond at Stanford SIEPR studied 5,179 customer support agents using a generative AI assistant. Average productivity rose 14 percent. The lowest-skilled quartile rose 34 percent. The highest-skilled quartile saw minimal lift.

Stanford SIEPR · Generative AI at Work · siepr.stanford.edu Stanford HAI · 2025 AI Index Report · hai.stanford.edu

2. Enterprise-level surveys report no productivity boost. A National Bureau of Economic Research survey of 6,000 executives across the US, UK, Germany, and Australia found 89 percent reporting no productivity impact from AI over the prior three years.

NBER survey · summarized in Fortune · fortune.com

3. Goldman projects 7 percent global GDP lift and 1.5 percentage points of productivity growth. Goldman Sachs Research’s headline AI forecast estimates roughly $7 trillion of additional global output over ten years, with 1.5 percentage points of additional productivity growth.

Goldman Sachs Research · Generative AI could raise global GDP by 7% · goldmansachs.com Goldman Sachs Research · AI is showing very positive signs of boosting GDP · goldmansachs.com

4. Enterprise AI spend is rising 65 percent per year, concentrated in agentic workflows. a16z’s 2025 survey of 100 enterprise CIOs found average LLM spend rising from $4.5M to $7M to a forecast $11.6M, with agentic inference the fastest-growing developer behavior.

a16z · How 100 Enterprise CIOs Are Building and Buying Gen AI in 2025 · a16z.com a16z · The Rise of Computer Use and Agentic Coworkers · a16z.com

5. 91 percent of executives expect agentic transformation. 13 percent have shipped it. Harvard Business Review’s 2026 agentic enterprise survey reports the gap between executive conviction and organizational readiness.

HBR · What Leadership Looks Like in an Agentic AI World · library.hbs.edu MIT Sloan Management Review · The Emerging Agentic Enterprise · sloanreview.mit.edu

6. Sequoia argues the next $1 trillion company will sell work, not software. Konstantine Buhler’s Agent Economy framework defines the three primitives of agent infrastructure: persistent identity, seamless communication protocols, and security and trust. Sequoia’s Services: The New Software essay extends the argument to a trillion-dollar profit pool.

Sequoia · Services: The New Software · quasa.io Sequoia · The Agent Economy · inferencebysequoia.substack.com

B · Complicating evidence

Three findings push back on the simple version of the story. They belong in any honest read of the institutional record.

METR finds AI tooling slows senior engineers by 20 percent on complex repositories. Model Evaluation and Threat Research’s 2025 study of working developers found AI tools added 20 percent to task time on average, inverting the Stanford pattern. The lift is not uniform across roles and complexity levels.

Thomson Reuters finds modest productivity gains for lawyers, accountants, and auditors. The 2025 report found targeted use of AI for document review and routine analysis produced gains that fell short of signaling widespread displacement.

HBR warns against treating agents as employees. A May 2026 HBR paper argues that the agent-as-employee analogy hides three things human workers bring that agents do not: stable context across the day, an instinct to escalate when something feels off, and accountability that survives a bad outcome. The recommendation is to run agent rollouts through systems-engineering templates with scoped permissions, observability, kill switches, and a clear owner per action.

C · Historical parallel

The current gap between task-level AI productivity studies and enterprise-level surveys looks remarkably like the Solow paradox of the late 1980s. Robert Solow, writing in the New York Times in 1987, observed that “you can see the computer age everywhere but in the productivity statistics.” The lag between visible technology and visible productivity stretched roughly a decade before the late-1990s acceleration. The operators who built the workflow during that lag captured the gains when the macro numbers finally turned.

D · Reading list

A starting point if you want to go deeper. Eight pieces, ordered by relevance to the three builds.

Stanford SIEPR · Generative AI at Work. The canonical task-level productivity study. Begin here.
Stanford HAI · 2025 AI Index Report. Comprehensive annual review. The productivity chapter is the relevant section.
a16z · How 100 Enterprise CIOs Are Building and Buying Gen AI in 2025. The cleanest read on enterprise spend and behavior.
a16z · The Rise of Computer Use and Agentic Coworkers. The agent thesis from the firm with the largest agentic portfolio.
HBR · What Leadership Looks Like in an Agentic AI World. The clearest executive-level framing of the adoption gap.
Sequoia · Services: The New Software. The trillion-dollar thesis.
Goldman Sachs Research · Generative AI could raise global GDP by 7%. The macro forecast.
MIT Sloan Management Review · The Emerging Agentic Enterprise. Four organizational tensions in agent adoption.

E · Source architecture note

The elite-research skill described in Build One is what produced this appendix. Every cited source sits inside the curated universe of Tier 1 publications, Tier 2 academic centers, and Tier 3 institutional research shops. The output format is a Research Brief. The full brief that anchors this appendix is published as a companion document.