Companies Are Becoming Tech Stacks. We Are All Becoming Companies.
The Operating Essay - A Weekly Reflection from Operating by John Brewton
Are you new to Operating?
You Might Enjoy These
(Access the entire library of articles and resources for $7.99 per month)
0 to 55,000 - The First 90 Days Playbook - What I’d Do Differently Knowing What I Know Now
The Operating by John Brewton Resource Library - No Filler. All Killer.
WINNING THE LOSER’S GAME: The Creator Economy and the Companies We All Need to Build
Become a paid subscriber to access all the content ($7.99 / month) or get to work with us and become a member of the Operating Founder cohort for $99 (annual).
Start Here
Most of us have spent our careers doing work that computers simply couldn’t do yet.
That sentence is worth sitting with.
Not because it’s provocative, but because it reframes so much; every org chart, every department, every management layer, every job title you’ve held or reported to. The entire apparatus of modern business has been, at its core, a system for organizing people to do cognitive and operational work that machines hadn’t yet learned to perform.
Management itself was the workaround. You needed managers because you needed people, and you needed people because the work couldn’t be done any other way.
Two weekends ago I was sitting on my couch working, my wife sitting across on our other couch doing the same. Something happened. After a 40 minute conversation and directional prompting session with Claude’s Opus 4.6, I watched the machine do twenty hours of my best work in twenty-five minutes.
I had tried equally rigorous exercises before, with various models, and they failed.
Not today.
Knowing something is coming and watching it arrive are different experiences entirely. The markets relative to SaaS certainly showed us this over the last two weeks.
This is an essay about two personal experiences, one from two years ago and one from the past two weeks. These are two stories about two distinct moments where I realized how obsolete my hard won skill sets had become.
We talk, conjecture, pontificate about all of this quite a lot.
I think our real stories are more useful. Hope this helps.
— j —
The Intern
Two years ago I hired a college student. Summer intern. Smart kid, willing to learn, zero Excel or business experience. To set the stage properly, he had never opened an Excel sheet in his life.
I gave him a premium ChatGPT membership and a stack of tasks.
The tasks were not simple. They were the kind of Excel work I’d spent two decades learning to do well:
Layered analysis
Conditional logic
Macro writing
Pivot tables on top of pivot tables
Reporting structures that pull from multiple data sources
Complex, clean, accurate output
The work required thinking in spreadsheets, not simply using them.
I spent twenty years building models, writing nested formulas, designing dashboards that could survive handoff to a client who’d never touch the underlying architecture. I learned this craft the way most operators my age did: slowly, painfully, one broken VLOOKUP at a time.
The intern learned it all in two days.
He did not master the theory or concepts. He learned to execute the output. He sat with ChatGPT and described what he needed, directed its writing of Python scripts, which were then placed into Google Script Editor, followed by downloads of the Google Sheets’ output as Excel files. He didn’t so much understand all the code. He didn’t need to. He understood the problem, described it clearly, and the machine built the solution.
By day three the quality of his output matched mine. By the end of the first week, some of it was better. Cleaner formatting. More elegant logic. Fewer redundancies in the cell architecture than I would have built by hand.
Then the competence compounded. Once he could build advanced spreadsheets, he could do data cleansing. Once he could cleanse data, he could build reporting pipelines. Within two weeks he was performing work that I would have previously needed to hire a specialist to do, or spend my own billable hours completing.
A college kid with no experience and a forty-dollar monthly subscription had replicated the functional output of a twenty-year practitioner.
I more than noticed. I didn’t file it away. It’s true that he couldn’t describe what needed building or articulate the more complex bits of the business logic and strategic intuition required for the larger project, but he could perform, with Chat, the most laborious parts of the work in profoundly less time.
That was the week I knew I was becoming obsolete.
Two Years Late, Saturday
I run Time-Driven Activity-Based Costing projects for consulting clients. If you haven’t encountered TDABC, the short version is this: it’s a rigorous methodology for understanding what things actually cost inside a business, not what an accounting system says they cost, but what they cost when you measure the time and resources consumed by every process, every activity, every transaction across a company.
These projects require extensive preparation. You need libraries of Excel workbooks (cost models, capacity calculations, process maps, time equations). You need interview documents tailored to every functional area of the business. You need intake forms calibrated to the client’s industry, their market position, their operational structure. Every document must conform to the TDABC framework while reflecting the specific characteristics of that particular company in that particular sector.
I’ve done two dozen of these projects over my career. Even with templates from prior engagements to build from, research and preparation takes twenty hours or more. The documents are complex. The spreadsheets contain thousands of cells governed by hundreds of interlocking equations. The interview guides require deep familiarity with the client’s industry. None of it is boilerplate.
Quick Note: Different than the story above, these were assets that needed to be aggressively engaged with by the client, using compounded, interlacing logic that was not reliably possible in the story I just told about two years prior.
Two Saturdays ago I sat down to prepare the document set for a new engagement.
I started with research. Forty-five minutes with Perplexity and SuperGrok, pulling industry data, market characteristics, sector-specific operational patterns for the client’s business. Then I opened a conversation with Claude’s Opus 4.6 and laid out the project. I explained the TDABC methodology in detail. I described every document I needed. I outlined the client’s industry context, the specific requirements for each workbook, the logic that would need to govern the equations. Importantly, I did not upload prior work of mine. I wanted to see what Opus could accomplish independently, and figured if that didn’t work, I could then upload my examples to help move things along.
Forty-five minutes of conversation. Ostensibly the briefing I would give a very capable analyst before sending them away for two days.
Claude built everything in under twenty-five minutes.
Every Excel workbook.
Every interview document.
Every intake form.
The spreadsheets contained thousands of cells. Over a hundred equations, all interconnected, all functional. The formatting was clean, polished, not just adequate, genuinely well-designed. The built-in logic worked. The math checked out. The documents reflected the client’s industry accurately.
I sat on my couch and stared at my laptop. Seventy minutes, start to finish. Research, briefing, and generation. Seventy minutes for what has reliably taken me fifteen to twenty hours across two dozen prior engagements.
The output wasn’t a rough draft. It was the finished product.
The Math
Here is what flooded my thinking in the next few hours.
I ran distribution companies. I ran manufacturing operations. I managed teams of inventory analysts, inside sales representatives, procurement specialists, sourcing managers. These were skilled people, in specialized departments, doing complex work: demand forecasting, vendor negotiation support, inventory optimization, cost analysis, purchase order management, customer account maintenance.
I started mentally calculating. If I were running those companies today, with these tools, what would the org chart look like?
Entire departments (inventory analysis, inside sales support, procurement operations, sourcing) could be reduced to single managers overseeing the full scope of work that previously required teams of five, eight, twelve people. A single analyst equipped with these platforms could produce the output of several dozen, based on the task structures and time allocations I knew intimately from years of managing those functions.
From my experience with the intern, I knew we were moving to this place, but it was in this moment that I understood the train had arrived at the station.
The Wall
There is a story that experienced operators tell ourselves. I told it to myself after watching the intern. The story goes like this:
The machine can execute, but it cannot think. It can build what you describe, but it cannot decide what needs building. The value is in the judgment, the strategy, the accumulated wisdom of having done this work for decades. The machine is a tool. We are the hands that wield it.
Sound familiar?
This story is partially true. Judgment still matters. Knowing what to ask for, and knowing whether the output is right, requires experience the machine doesn’t have and the intern hadn’t yet developed. I spent forty-five minutes briefing Claude because I’ve done twenty-four of these projects. A novice couldn’t have written that brief. The machine needed my expertise to produce the work that replicated my expertise.
But partial truths are dangerous. Comfort is dangerous. The wall between “can execute” and “can decide” is not a wall. It’s a membrane. And it’s thinning.
The intern didn’t just learn to execute my tasks. He started to understand why the tasks existed. Given enough reps with these tools, he will develop judgment, not in twenty years, the way I did, but in two or three… maybe only one?
The tools are compressing the learning curve alongside the labor.
That is the part most haven’t processed yet. It’s not just that AI does the work faster. It’s that AI makes the learning faster. The experience gap that took decades to build can now be closed in months. It won’t be eliminated entirely, but enough so that the lower cost of the less experienced person, or the cost of the model having enough compute, will be more appealing to the company than paying for your experience.
Not eliminated. Close enough. Better enough. Good enough.
That’s a critical distinction we don’t mention with appropriate frequency. It’s not that the models have to do the work better than we can. They just have to be good enough for the cost savings to trump our value in the eyes of a company working to control costs and become more price competitive. And as markets become commoditized, the race to zero around price heats up like a mid-July Tuesday in South Texas. Also, for anyone who hasn’t noticed, the velocity at which markets are becoming commoditized, given these new tools, is extraordinary.
Scary…exciting…scary exciting.
Seventy Minutes
This is not a warning. The time for warnings has long since passed.
If you’ve done the kind of work I’ve done, you’re probably becoming obsolete. If you’ve been laid off, the next corporate job might not be out there.
At least not in the form you remember it.
So what do we do?
Go back to where we started. We have spent our careers doing work that computers couldn’t do yet. And the business of management existed to coordinate the people doing that work. Both layers are compressing. The work is being absorbed by the machines. The management of the work is being absorbed alongside it. Try talking to Gemini 3 about management problems, tactics and strategy. It’s an excellent advisor. Companies are becoming tech stacks with a thin layer of human judgment placed on top.
But here’s the good news…
The same collapse that makes you obsolete inside a company makes you viable as one.
Think about what a company actually is. It’s a bundle of capabilities (analysis, operations, communication, production, distribution) organized under one entity and deployed against a market. For the entirety of modern business, assembling that bundle required capital, headcount, and infrastructure.
It required people doing work that computers couldn’t do yet.
That requirement is dissolving. The tools that allow a corporation to replace a twelve-person department with a single manager are the same tools that allow a single person to operate with the capability of a department.
The cost of capability is collapsing.
Not partially.
As we have understood capability, it will collapse entirely.
Twenty years of expertise, replicated in seventy minutes. That’s the threat if you’re sitting inside a company waiting for the org chart to protect you.
It’s the opportunity if you’re willing to become the company yourself.
I don’t mean this in the motivational, hustle-culture sense of “everyone should be an entrepreneur.” I mean it structurally. The economics now favor the individual who can direct these tools across a full stack of business functions (analysis, content, operations, client delivery, brand) over the employee who performs one function inside an organization that is actively looking to automate that function away.
The essential skill of the post-AI economy is not prompting.
It’s not “learning to work with AI.”
Those are table stakes. The essential skill is the ability to build and operate a business. To take the same technology that is being used to make you obsolete and deploy it as your own infrastructure. To become, in effect, a company of one with the output of many.
Companies are becoming tech stacks. We are all becoming companies.
Twenty years of expertise. Seventy minutes. Output that was equal or better.
That is the fact for all of us. What you build with it is yours to decide.
- j -
Are you new to Operating?
You Might Enjoy These
(Access the entire library of articles and resources for $7.99 per month)
0 to 55,000 - The First 90 Days Playbook - What I’d Do Differently Knowing What I Know Now
The Operating by John Brewton Resource Library - No Filler. All Killer.
WINNING THE LOSER’S GAME: The Creator Economy and the Companies We All Need to Build
Get to work with us and become a member of the Operating Founder cohort for $99.
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? He still cringes at his early LinkedIn posts and loves making content each and everyday, despite the protestations of his beloved wife, Fabiola, at times.
This is a fascinating perspective on the evolution of business structures. The concept of companies becoming modular tech stacks really resonates: it mirrors how individual professionals are also building their own personal tech stacks to operate more efficiently. Great insights!
Brilliant post as ever John! In compressing the learning curve do you think that there might be a risk that companies opt to skip it altogether rather than use it as an amazing opportunity for genuine innovation?