Start Here

Every digital platform you use is a marketplace.

LinkedIn, YouTube, Substack, Amazon, Google, TikTok, eBay, Instagram, etc.

Each platform exists to solve a coordination problem.

Economists call these n-sided markets.

You probably call them your “content strategy.”

The moment you understand that on each of these platforms you are participating in a market governed by specific pricing mechanics, information asymmetries, and power structures, you stop subsidizing their platform and start negotiating your position inside an economic system.

That is a different game entirely.

Thrilled to have collaborated on this article with my esteemed friend and colleague, Jonas Braadbaart:

Jonas Helps business leaders become AI operators. He is an AI architect and AI agency owner with 10+ years of experience designing, building and delivering AI solutions. Currently, he is building an AI film making studio:

https://www.project-skia.com/

Here is a link to a prior Substack Lives we enjoyed doing together:

1. Creator Economics and the N-Sided Marketplace

2. What Works on Substack Will Get You Ignored on LinkedIn (And Vice Versa)

The Economics Governing Your Platform Earnings Potential

In 2003, Rochet and Tirole published the paper that formalized how platforms actually set prices. Their central insight was deceptively simple:

In a two-sided market, the structure of prices matters more than the total price level. A platform does not just charge what the market will bear.

It strategically subsidizes one side to attract the other:

• Substack charges readers nothing and takes 10% from writers.

• Amazon charges buyers nothing for Prime search results and takes 50% from sellers.

• Google charges searchers nothing and monetizes publishers’ traffic into ad revenue.

• Youtube Shorts earn next to nothing compared to long-form content published on the platform.

The pattern is identical every time.

One side is the product. The other side is the customer.

If you are not sure which one you are, check how you are paying.

Three years later, Mark Armstrong identified something darker.

He called it the “competitive bottleneck.”

When one side of a market “single-homes” (readers who use one app, buyers who search one store) and the other side multi-homes (creators who post everywhere, sellers who list everywhere), the platform develops a local monopoly over access to the single-homing side.

It can then price the multi-homing side as aggressively as it wants.

Armstrong’s Proposition 4 states, almost clinically, that platforms will systematically under-invest in the experience of the multi-homing side. They have nowhere else to go to reach that particular audience.

This is the market’s structural equilibrium.

As a creator, you are typically on the multi-homing side (notable exceptions include Joe Rogan’s USD 100 million Spotify deal, but those are very rare indeed).

Just as new investors should assume they are not Warren Buffett, it would be prudent to assume here that you are not Joe Rogan.

The Algorithm Is the Marketplace

Susan Athey’s work at Stanford reframes platforms not as transaction venues but as information engines.

The real service is not hosting your content. It is reducing the information asymmetry between you and the person who might want to read, watch, or buy from you.

Reviews, ratings, recommendation algorithms, verification badges…

These are the mechanisms preventing the market from collapsing into what Akerlof called the “lemons problem,” where low quality drives out high.

This is why algorithmic change feels existential, because it actually is for most creators. The algorithm is the marketplace’s allocation mechanism for attention. When LinkedIn replaced thousands of fragmented ranking models with a single 150-billion-parameter foundation model that reads your content, evaluates your profile, and matches you to audiences, it did not update a feature.

It redesigned the market’s central pricing mechanism.

The Result: LinkedIn company page reach dropped 60 to 66% between 2024 and early 2026. Personal profiles now receive roughly 561% more distribution. The platform decided that corporate broadcasting is less valuable than human expertise.

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

Operating Stories: Hermès and the Manufacturing of Scarcity

Operating Economics: Building Antifragile Companies

Become a paid subscriber to access all the content ($7.99 / month)

Subscribe now

The Four Seasons of Getting Squeezed

Every platform follows the same extraction lifecycle:

Phase One–Early Growth: The platform subsidizes both sides. PayPal paid users $10 to sign up. Substack recruited journalists with guaranteed advances. TikTok handed virality to unknowns like confetti.

Phase Two–Late Growth: Network effects compound. Users become dependent. Switching costs rise. Not because leaving is technically hard, but because your audience, your reviews, your history, and your data all live behind someone else’s login screen.

Phase Three–Early Maturity: Monetization tightens. Take rates climb. Organic reach quietly declines, rebranded as “algorithm improvement.” Paid promotion becomes the price of visibility you once earned for free.

Phase Four–Active Extraction. Amazon’s effective seller take rate has crossed 50%. Google’s AI Overviews have driven organic click-through rates down 61%, with 58.5% of U.S. searches now ending in zero clicks. The platform answers the question itself. The publisher who created the information maybe gets a citation, but is not guaranteed traffic.

In general, the best time for getting onto any platform is in phase one or phase two of their lifecycle. You’re not fighting the platform.

Competition is light.

Attention is heavily subsidized.

In phases three and four, achieving ROI (return on investment) on your position will get harder and harder. Platforms become pay-to-play.

Think about Amazon or Google as rent-extracting monopolies.

Three Examples

If all of this still sounds very abstract, don’t worry, we’ve got you covered.

Let’s look at three platforms in more detail:

LinkedIn–multi-sided professional platform (6+ sides):

LinkedIn is one of the strongest “information engines” out there in terms of real identity, verifiable employment history, and peer endorsements. Their 360Brew algorithm (rolled out gradually from March 2025) scans all text (headlines, about sections, captions, comments) for meaning. It prioritizes “authentic engagement” and industry-specific insights over generic viral posts.

This was the most significant algorithm change in LinkedIn’s history.

Phase: Early Maturity: The platform algorithms have started suppressing organic reach for companies → to force ad spend → and selectively reward top creators to maintain content quality → then gradually expand monetization to mid-tier creators.

These changes, coupled with a feed composition shift (11% ads, 31% Top Creator content) means the average user’s organic reach is structurally declining.

Get started today or lose out on the platform.

Substack–two-sided creator platform evolving toward multi-sided (3–4 sides):

Substack’s information engine has grown significantly these last two years.

Data from 2025 showed the recommendation network drove 40–50% of all new subscriptions and 25% of new paid subscriptions, and is still the platform’s most powerful discovery mechanism. 40% of all subscriptions now come from inside the Substack network (app + recommendations).

This means Substack is becoming a genuine discovery platform, not just an email delivery tool, which both strengthens its value to writers and deepens its competitive bottleneck.

Phase: Late Growth: the best creator economics in the market, but the first cracks are appearing

Amazon–N-sided commerce platform (6+ sides):

Amazon’s information engine is undergoing its biggest transformation since reviews were introduced.

The Rufus AI assistant and COSMO semantic knowledge graph mean sellers now optimize for three distinct systems simultaneously: the A9/A10 keyword algorithm, COSMO’s semantic understanding, and Rufus’s conversational layer.

This dramatically increases the opacity of the matching engine and the resulting information asymmetry between Amazon and its sellers–allowing Amazon to smartly position their own products and brands.

Phase: Active Extraction - One of the oldest platforms, and not entirely coincidentally, also one of the most advanced cases of rent-seeking in big tech.

download high res pdf here

The High Over, March 2026

Best positions (invest heavily)

1. Substack — Late Growth phase, lowest take rate (10%), subscriber list portable, discovery engine strengthening (40%+ of subs from in-app). The window is now before monetization pressure from the $1.1B valuation creates extraction incentives.

2. YouTube (long-form) — Strong bottleneck works FOR you if you’re established. Early Maturity but stable economics. 55% revenue share is the best rate in video at scale. The Shorts format shift is extractive, but long-form remains the gold standard for creator revenue.

3. LinkedIn — Early Maturity means the platform is still investing in creators via BrandLink. 360Brew algorithm rewards niche expertise. Its core professional audience is high-value. But the organic reach suppression for company pages signals the direction, personal brand investment is time-sensitive as the platform moves towards the active extraction stage.

Maintain presence (don’t over-invest)

4. TikTok — Regulatory risk resolved (Jan 2026 USDS deal). Great for reach/discovery with improved Creator Rewards ($0.40–$1.00/1K views, up 20x). Its new follower-first algorithm changes are ending the “anyone can go viral” era. Use as a top-of-funnel. TikTok Shop commerce is still a great opportunity in 2026, despite increasing take rates 4% -> 6% -> 8%.

5. YouTube Shorts — Necessary for discovery, but structurally extractive (55% platform take, 50–100x lower RPM than long-form). Feed long-form. Don’t use the channel as a standalone pillar in your content strategy–only for discovery.

6. Instagram — Mature and strategically confused but 3B MAU makes it impossible to ignore. Commerce retreat means it’s an advertising platform, not a marketplace. Use for brand building in visual niches.

7. eBay — If you sell collectibles, pre-owned, or parts, this is experiencing a genuine renaissance (focus categories +16% in Q4 2025). AI features are a real competitive advantage. Otherwise, Amazon dominates.

Proceed with caution

8. Amazon (as seller) — Active Extraction: 45–51% effective take rate and rising. Be there because you have to, but build direct channels aggressively. The FTC trial (October 2026) is the single most important event for marketplace sellers this decade — plan for both outcomes.

9. Google Search (as publisher) — Active Extraction via AI Overviews. Organic CTR drops 61% when AI Overviews appear. Plan for 40%+ traffic decline over 3 years. Diversify to email, social, and direct urgently.

Exit or minimize

10. X (Twitter) — Declining platform (Threads overtook X on mobile DAU, Jan 2026). User base shrinking 10–15% YoY. Ad revenue recovering from a deeply depressed base but CPMs still 70% below pre-Musk levels. Maintain a minimal presence; invest content-creation time on LinkedIn, Substack, YouTube, or Threads. Worth noting: Communities around technology (AI engineering), venture capital, crypto, and the investing community maintain a strong presence on the platform, making presence an imperative if you want to build awareness for your work in these important sectors.

The Power Law You Are Living Inside

A 2025 study analyzing 104,000+ Patreon creators found that earnings across platforms follow a power law distribution with an exponent around α = 2.0. The mechanism generating creator income online behaves more like capital accumulation than labor income.

Initial advantage compounds.

Algorithmic recommendation accelerates the compounding.

The finding that should concern you: Platforms with stronger recommendation engines exhibit more inequality and lower median earnings. Platforms with weaker algorithmic curation show higher median incomes. The algorithm does not just pick winners. It mathematically suppresses the middle class.

Recent data confirms the trend is accelerating.

The top 10% of creators earned 62% of total payments in 2025, up from 53% in 2023. Median creator earnings declined from $3,500 to $3,000. More creators. More money. Less of it reaching most of them.

Three Moves To Strengthen Your Business Today

Stop thinking like a creator. Start thinking like a strategist in markets with known structural properties.

Build portability. Your email list is the only asset the platform cannot reprice, throttle, or algorithmically suppress. Every interaction on a platform you do not own should have one objective: move the relationship to a channel you do own. Everything else is rented reach.

Layer revenue. If more than 30% of your income depends on any single platform’s algorithm or fee structure, you are not building a business. You are being allocated income by someone else’s information engine. That allocation can change on a Tuesday afternoon with no notice.

Read the lifecycle. When organic reach declines and paid promotion becomes necessary to maintain prior visibility, you are entering Phase Three. When the platform launches its own competing product or its AI begins answering questions with your content without sending you traffic, you are in Phase Four. These are not abstract signals. They are measurable. Monitor take rates, reach trends, and new fee introductions like the operating metrics they are.

The platforms will keep evolving. They need to.

And N-sided markets solve real coordination problems and create genuine economic value. For example, a 2019 study by Brynjolfsson et al found that the median US consumer would have been willing to pay up to $17,530 per year for access to search. Google’s US Search ad revenue that year was roughly $45–47 billion across 150 million users, or about $300 per user.

So in this estimate, in 2019, with Google Search in the late growth stage of its lifecycle, Google captured roughly 2% of the value created by its n-sided marketplace.

In 2026, with Google in active extraction mode, revenue per user has roughly doubled to ~$675 (still under 4%) but organic click-through rates have been halved by AI Overviews. Google isn’t taking a bigger slice of a shared pie, but it is absorbing more and more of the value that used to flow to content publishers, historically, one of the key participants in their marketplace.

Final Thought

The economics that make platforms powerful are the same economics that make participants vulnerable.

The difference between being extracted from and being strategically positioned is knowing which side of the equation you are standing on.

Every platform you build on today was designed to solve someone else's coordination problem. Not yours.

The economics are not hidden. They are published in quarterly earnings calls, buried in fee schedule updates, and visible in every percentage point of reach you lost last year. The only question is whether you read them like an economic analyst or experience them as the commodity you can easily become.

Companies are becoming tech stacks.

We are all becoming companies.

Act accordingly.

Hope this helped.

- John & Jonas -

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.

Every week, I bring in one amazing friend and expert I trust to share hard-won lessons from the trenches of their work worlds. This week, my friend Jens Stark is in the chair. He brings 30,000+ hours of experience in post-sales environments to a topic most founders treat as an afterthought until there’s a five-alarm fire: customer care.

Jens runs Scaling Customer Value on Substack, focused squarely on monetizing and engaging the customers you already have. If you're burning cash on acquisition but ignoring the people who already said yes, this one's going to recalibrate your thinking.

Pay attention, and go subscribe to Jens's newsletter here when you're done.

- j -

Subscribe now

Why complaints are a good thing

Most companies think of complaints as something negative and try to avoid them. Whilst it is certainly a good idea to keep customers happy, what is less obvious is that complaints are a goldmine of information for company operators.

The most honest and raw feedback you will receive from customers can be found in your complaints department. Valuable information and customer insights are hidden within the contents of the customer communications and the complaints data.

If the only people who see this information are your company’s complaint handlers, you are missing out on a wider opportunity.

Having an effective operational process to manage complaints is important, but creating systems and an operating structure to enable you to review, analyze and discuss the customer feedback and insights at a business leadership level is how you enable change.

“Your most dissatisfied customers are your greatest source of learning.”

— Bill Gates

—--------------------------------------------------

You might have come across the complaints iceberg concept before, which suggests that complaints represent only the visible tip of the iceberg. Hidden underneath the surface is a much larger volume of unhappy customers who haven’t voiced their concerns.

In practice, less than 5% of the dissatisfied customers might go on to file a formal complaint. From the vast majority of customers that won’t raise any concerns, some will quietly stop using your product while others will start preparing to leave or switch to a competitor.

This is called silent attrition and it can be hard to spot and difficult to deal with. Once your business realises that there is a risk of losing the customer, the decision to move away from your company’s products and services has already been made.

Because of these dynamics, you can scale your business impact by approaching complaint handling proactively and strategically. When you stop thinking about complaints as items on a conveyor belt, the real question isn’t about processing volumes, speed or efficiency – it’s about the steps and actions you need to take to prevent them from re-appearing there the next time:

1. Creating a categorisation of different complaint themes

2. Measuring volumes, proportions and trendlines

3. Identifying root-causes of complaints

4. Customer feedback & complaint reporting loop to business decision-makers

5. Working collaboratively to mitigate the root causes of complaints

The benefits of reducing complaint volumes are not just operational but also financial. Lower complaint volumes means less operational cost and redress.

In addition, the continuous improvements that follow from the above methodology means having fewer weak points – across your products, policies, processes and people.

What do you reckon are the chances that such improvements will enable you to command a higher price point for your products & services in the future?

In one of the first business textbooks that I came across when I first started studying business some two decades ago, I vividly recall reading the following statement:

• A happy customer will tell 5 people about their experience

• An unhappy customer will tell 20 people about their experience

This is insightful because it highlights the importance of positive customer experiences and also why nipping those complaints in the bud quickly can help limit word-of-mouth conversations about a negative experience with your brand.

However, technological advances over time have radically changed how people communicate with each other. You now need to consider the digitisation of complaints; they can be emailed to your business with other people on copy, your response to the complaint can be forwarded to someone else with the click of a button, or the complaint could be made public on social media.

Possibly the world’s most impressive complaint about poor customer service comes from Canadian musician Dave Carroll and his band, Sons of Maxwell. “United Breaks Guitars” is a protest song which retells the story of how his guitar was broken during a trip with United Airlines in 2008 and the uncooperative reaction from the airline.

Carroll posted on YouTube on July 6, 2009; where it amassed 150k views in one day, prompting the airline to contact the musician. It was also reported that within a month of the video being posted online, the company’s stock price fell 10% and wiped out shareholder value of $180 million. 17 years on, the video is still going strong with 30 million views, 934k likes and 62k comments!

The stakes of complaint handling are higher than ever before and all of this is a very good reason for why you should consider bumping up complaint management in your list of business priorities.

By approaching complaints proactively and working through them holistically and methodically, you will be able to extract valuable customer insights and execute customer-needs-driven changes that will improve your business and make it more resilient.

- Jens -

Go subscribe to Jens’s newsletter here!

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

• Operating Stories: Hermès and the Manufacturing of Scarcity

• Operating Economics: Building Antifragile Companies

• How Three Decades Rewrote the Economics of Companies

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).

[Join the Operating Founder Program — 40 spots remaining]

Subscribe now

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.

• The Sorting Year: What We Learned From 2025 & Where We’re Going in 2026

• Operating Economics: The Optimization Company

• Operating Economics: Building Antifragile Companies

• Start With The Work, Not The Why

• Sorry Is Not A Solution

• How Three Decades Rewrote the Economics of Companies

• We Are All Becoming Operators

Subscribe now

Welcome Readers! This article is the first of a two-part series, a collaboration between Operating by John Brewton and The Data Letter author and Asaura AI founder, Hodman Murad.

Hodman Murad founded Asaura AI to develop artificial intelligence systems that help people with ADHD and executive dysfunction. She publishes Asaura, exploring how AI can enhance human productivity and mental health. She builds in public, connecting high-performance systems with their users.

Hodman also writes The Data Letter, analyzing common failures in AI and data systems, including hallucinated growth metrics, customer loss from misguided optimization, and mismanaged AI agents.

We set out to merge my expertise in corporate financial and economic analysis with Hodman’s deep technical background in data science and engineering, with the hope of creating a case study valuable to operators, founders, builders, students, and investors alike. JPMorgan Chase’s comprehensive approach to AI transformation stands out as unparalleled in both scope and execution. Amid widespread narratives about struggling digital transformations, we wanted to examine what’s demonstrably working.

Hope this helps! Thank you for reading!

Part Two from Hodman, How JPMorgan Built AI Infrastructure That Delivers $2 Billion in Annual Returns, will be released on Wednesday of this week, and we will also be doing a Substack Live to talk about this work.

Start Here: Executive Summary

JPMorgan Chase has executed one of the most aggressive and transparent artificial intelligence strategies in global banking, transforming a $15 billion technology budget in 2021 into an $18 billion AI-enabled competitive engine by 2025. This analysis examines the bank’s financial performance since 2021 alongside explicit mentions of AI spending and operational impact across quarterly earnings reports, annual shareholder letters, and executive statements. The evidence reveals a deliberate, multi-year AI buildout that has delivered measurable returns—from $100 million in business value in 2022 to $2 billion annually by 2025—while positioning the institution as the clear AI leader among global financial institutions.

JPMorgan Chase delivered strong revenue growth from 2021-2024, with managed revenue increasing from $127.1B to $180.6B (42% growth). Despite a dip in 2022 (post-reserve release period), net income reached a record $58.5B in 2024, while ROTCE remained consistently strong between 18-23%.

I. Financial Performance Overview: Resilience Through Volatility (2021-2024)

JPMorgan Chase has delivered exceptional financial performance throughout a period marked by pandemic recovery, reserve release volatility, inflation, regional banking turmoil, and aggressive Federal Reserve tightening. The firm’s results demonstrate both earnings power and the value of strategic technology investment.

I’m building something different in 2026: Personally working 1:1 with just 100 serious operators, founders, creators, and builders.

4 quarterly 60-minute working sessions with me (recorded for you) to build your operating strategy, crush team issues, sharpen your brand, or fix whatever’s blocking growth.

For the first 25 who join, I’ve slashed the Operating Founder tier from $550/year to $99/year. Once those spots fill, the price jumps back.

Why $99? I want to partner closely with 100 inventive builders in our AI-first era, learn your stories, and see what’s really working.

Subscribe now

Annual Performance Trajectory (2021–2024)

JPMorgan Chase achieved robust revenue growth from 2021 to 2024, with managed revenue rising 42% from $127.1 billion to a record $180.6 billion. Net income reached an all-time high of $58.5 billion in 2024, while ROTCE remained consistently strong in the 18–23% range.

2021: Reserve Releases and Capital Markets Strength
Reported net income of $48.3 billion and 23% ROTCE benefited from ~$9.2 billion in pandemic-era credit reserve releases. Normalized for these releases, earnings approximated $39.1 billion with 18.5% ROTCE—still robust. Results were bolstered by elevated investment banking fees (e.g., Q1 up 222% YoY) and strong trading revenue amid record capital markets activity.

Quarterly EPS volatility from 2021-2024 reflects the transition from pandemic reserve releases ($4.50 in Q1 2021) through normalization (2022 lows around $2.63-$2.76) to record sustained profitability in 2024. The Q2 2024 spike to $6.12 represents peak quarterly earnings performance.

2022: Normalization and NII Tailwinds
Revenue reached a then-record $132.3 billion, but net income fell to $37.7 billion (18% ROTCE) as reserve builds resumed amid economic uncertainty. Rising interest rates drove significant NII growth (e.g., +72% YoY ex-markets in Q4), offsetting weaker investment banking volumes. The firm retained its #1 global IB fee ranking (8% share) and reaffirmed its 17% through-the-cycle ROTCE target.

2023: Resilience Amid Banking Stress
Despite the regional banking crisis, revenue hit a sixth consecutive record of $162.4 billion, with net income of $49.6 billion and 21% ROTCE. The FDIC-assisted First Republic acquisition in May enhanced the franchise and stability. Core performance remained solid (e.g., normalized Q4 ROTCE ~19% excluding FDIC assessment and securities losses), with meaningful market share gains in consumer and investment banking.

2024: Record Profitability
Managed revenue climbed to $180.6 billion, with net income of $58.5 billion (diluted EPS $19.79, +21.8% YoY) and 20% ROTCE. Quarterly results showed consistency and strength, culminating in a Q4 EPS beat ($4.81 vs. $4.03 expected) driven by broad-based momentum across consumer banking, investment banking, and asset/wealth management.

Throughout the period, JPMorgan sustained or expanded leadership positions, ranking #1 globally in investment banking fees (9.1% share in 2024), Markets revenue, retail deposits (11.3% share), and credit card sales/outstandings—reflecting durable competitive advantages and operational excellence.

Multi-Year Financial Perspective

From 2021 through 2024, JPMorgan increased managed revenue by 42%—from $127.1 billion to $180.6 billion. Even accounting for the 2021 reserve release distortions, core profitability strengthened materially. The bank’s fortress balance sheet metrics remained robust throughout: CET1 capital ratios stayed well above 13%, tangible common equity grew from $203 billion (2021) to $230 billion (2024), and liquid assets exceeded $1.4 trillion.​

Importantly, this financial performance occurred alongside significant technology investment acceleration, which we examine in detail below.

II. The Architecture of AI Investment: A Five-Year Buildout (2021-2026)

JPMorgan Chase’s AI strategy represents a systematic, multi-billion-dollar transformation that predates the November 2022 launch of ChatGPT. The bank had been deploying predictive AI and machine learning in production since at least 2014, giving it a decade-long head start when generative AI emerged as a new frontier.​

Technology Budget Evolution and Composition

JPMorgan Chase’s AI investment demonstrates accelerating returns: technology budget increased 20% from 2022-2026 ($15B to $18B), while AI-delivered business value surged from $100M to $2B annually. Use cases are projected to more than double from 450 to 1,000 by end of 2026.

JPMorgan's infrastructure transformation accelerated meaningfully from 2022-2025: cloud adoption increased 71% (38% to 65% of applications) while modern infrastructure penetration reached 80%. This foundation enables the bank's AI ambitions—generative AI and large language models require cloud-scale compute and modern data architectures.

The bank’s technology spending has increased steadily and strategically from approximately $15 billion in 2021-2022 to $18 billion by 2025, representing roughly 10% of annual revenue, a level Dimon characterizes as “less than most other companies”. This budget encompasses far more than AI alone, including modernization, cybersecurity, data centers, public cloud migration, and product development. However, AI has become an increasingly central component and strategic priority.​

JPMorgan's 2023 technology allocation prioritized customer-facing innovation ($4.0B) and platform modernization ($3.2B), while engineering improvements generated $500M in offsetting savings. This strategic balance between growth investment and efficiency optimization characterizes the bank's approach to technology spending.

In 2023, CIO Lori Beer detailed the allocation at Investor Day: $7.2 billion was directed to new technology investments, split between $4 billion for customer/client experiences and product development, and $3.2 billion for firm-wide platforms including development, data, public cloud, and cybersecurity infrastructure. Notably, modernization investments delivered $300 million in savings through engineering practices and $200 million through infrastructure optimization—demonstrating that technology spending generates immediate efficiency returns even before AI-specific gains.​

By 2024, the dedicated AI investment reached approximately $2 billion annually—roughly 11% of the total technology budget. This figure, confirmed by CEO Jamie Dimon in October 2025, represents direct AI-related spending on talent, infrastructure, model development, and deployment rather than the broader technology envelope.​

JPMorgan Chase's AI investment demonstrates accelerating returns: technology budget increased 20% from 2022-2026 ($15B to $18B), while AI-delivered business value surged from $100M to $2B annually. Use cases are projected to more than double from 450 to 1,000 by end of 2026.

AI Talent and Organizational Infrastructure

JPMorgan has built the largest AI organization in global banking, a critical competitive advantage that enables execution at scale. As of the 2023 shareholder letter, the bank employed over 2,000 AI and machine learning experts and data scientists. The 2023 Investor Day provided greater granularity: more than 900 data scientists, 600 machine learning engineers, approximately 1,000 people in data management, and a 200-person AI research team focused on “the hardest problems and new frontiers of finance”.​

These numbers continued to grow through 2024-2025. By 2025, the bank was hiring specialists such as Sri Shivananda (former PayPal CTO) as firmwide CTO and Manoj Sindhwani (formerly Amazon Alexa) as CIO of the Chief Data and Analytics Office—both hired explicitly to accelerate AI adoption. In June 2023, the bank created a new C-suite position, Chief Data and Analytics Officer, appointing Teresa Heitsenrether to the role, and added the position to the Operating Committee, a clear signal of AI’s strategic priority.​

Rather than centralizing all AI work, JPMorgan distributed AI/ML talent across lines of business while maintaining centralized platforms, governance, and research capabilities. This enables rapid experimentation and deployment tailored to specific business needs (credit cards, trading desks, wealth management) while ensuring consistent risk management and compliance standards.​

JPMorgan expanded its AI workforce by 67% from 2022-2025 (1,500 to 2,500+ specialists) while production use cases grew 50% (300 to 450). The workforce growth acceleration in 2024-2025 positions the bank for its goal of reaching 1,000 use cases by 2026.

Use Case Proliferation and Production Deployment

The bank’s AI strategy is distinguished by its focus on production deployment rather than pilots. This operational discipline creates measurable business value and differentiates JPMorgan from competitors still experimenting with proofs of concept.

2021-2022: Establishing the Foundation

By 2022, JPMorgan had more than 300 AI use cases in production, deployed across risk management, prospecting, marketing, customer experience, and fraud prevention. These applications ran throughout the bank’s global payments processing and money movement systems, handling trillions of dollars in transactions daily. The 2022 shareholder letter emphasized that “AI has already added significant value,” specifically citing “significantly decreased risk in our retail business (by reducing fraud and illicit activity) and improved trading optimization and portfolio construction”.​

2023: Acceleration and Generative AI Entry

The number of use cases increased 34% year-over-year to over 400 by mid-2023. The 2023 Investor Day in May—six months after ChatGPT’s launch—marked a strategic inflection point. CIO Lori Beer announced the bank was “actively evaluating opportunities with large language models and see great potential in that space”. This signaled JPMorgan’s entry into generative AI experimentation while continuing to scale proven predictive AI applications.​

The bank ranked #1 on Evident AI’s Index, the first public benchmark measuring AI maturity across major banks. This ranking validated JPMorgan’s operational lead: while other institutions were announcing AI intentions, JPMorgan was delivering production results at scale.​

2024: Generative AI in Production

By April 2024, Jamie Dimon’s shareholder letter reported over 400 use cases in production spanning fraud prevention, marketing, customer service, software engineering, operations, and employee productivity. The bank had begun deploying generative AI in customer-facing and operational roles, moving beyond internal efficiency to client experience enhancement.​

The LLM Suite—JPMorgan’s proprietary large language model platform—launched in 2024 and rapidly scaled to 200,000 employees (63% of workforce) using it regularly. Applications included research summarization, contract scanning, code generation, and client intelligence. Investment bankers were automating 40% of research tasks; portfolio managers cut research time by up to 83%.​

2025-2026: Agentic AI and Workflow Transformation

By late 2025, JPMorgan had approximately 450 AI use cases in production. However, the bank’s ambition expanded significantly: CEO Dimon stated in early 2025 that he anticipated reaching 1,000 use cases “by next year”—more than doubling the portfolio within 12 months.​

The nature of applications was also evolving. The bank deployed “agentic AI”—autonomous systems capable of executing multi-step tasks without human intervention. Examples included AI agents that create investment banking presentations in 30 seconds, draft confidential M&A memos, and manage complex back-office processes like trade settlement and account setup.​

In January 2026, JPMorgan announced Proxy IQ, an in-house AI platform that replaced external human proxy advisors for shareholder voting in the U.S., a striking demonstration of AI taking over complex governance decisions.​

III. Quantifying Returns: From $100M to $2B in Annual Business Value

JPMorgan's AI business value surged 20x from $100M (2022) to $2.0B (2025-2026), with the steepest acceleration in 2023 when the bank raised its target from $1B to $1.5B and exceeded it. The 2025-2026 plateau at $2B reflects mature use cases delivering consistent returns while new initiatives scale.

I’m building something different in 2026: Personally working 1:1 with just 100 serious operators, founders, creators, and builders.

4 quarterly 60-minute working sessions with me (recorded for you) to build your operating strategy, crush team issues, sharpen your brand, or fix whatever’s blocking growth.

For the first 25 who join, I’ve slashed the Operating Founder tier from $550/year to $99/year. Once those spots fill, the price jumps back.

Why $99? I want to partner closely with 100 inventive builders in our AI-first era, learn your stories, and see what’s really working.

Subscribe now

The bank has been unusually transparent about the financial returns from AI investment, a practice that distinguishes it from competitors offering only qualitative claims. This transparency reflects confidence in measurement rigor and creates accountability for continued investment.

The Measurement Framework

Unlike many firms that tout “productivity improvements” without financial quantification, JPMorgan tracks AI returns through specific business value metrics: revenue uplift, expense reduction, cost avoidance, and risk mitigation. CFO Jeremy Barnum emphasized in January 2024 that the bank takes a “pragmatic and disciplined approach” focused on “actual results” rather than “chasing shiny objects”.​

The bank does not attempt to measure every productivity gain at a granular level. As Barnum explained on the Q3 2025 earnings call, mandating precise ROI measurement for every AI application can be counterproductive, pushing teams to “use AI in ways that are actually not efficient and that distract you from doing the underlying process reengineering that you need to do”. Instead, JPMorgan measures high-impact use cases while accepting that broader productivity tailwinds are occurring even if not precisely captured.​

Documented Returns by Year

2022: Early Returns ($100M)

At the 2023 Investor Day, CIO Lori Beer disclosed that AI-driven growth signals and product suggestions in the Commercial Bank had delivered $100 million of benefit in 2022. This single use case—providing bankers with AI-powered client insights—demonstrated material value from a narrow application.​

2023: Accelerating Value ($1.5B)

In May 2023, JPMorgan raised its annual AI business value target from $1 billion to $1.5 billion for the year, with Beer stating the bank was “ahead of plan” on the original commitment. By year-end, the bank had achieved this target.​

Specific contributors included:

• Retail personalization: AI-driven credit card upgrade offers and product recommendations delivered over $220 million in benefit during the 12 months ending mid-2023​

• Commercial banking sales intelligence: $100 million in benefit from AI-powered growth signals and product suggestions to bankers​

• Fraud prevention: Significant cost avoidance, though specific figures not disclosed in 2023

The bank also reported that AI ROI increased 25% from 2021 to 2022, demonstrating improving returns as use cases matured and the organization became more sophisticated in AI deployment.​

2024: Scaling Returns ($1.5B-$2.0B)

President and COO Daniel Pinto stated in 2024 that AI use cases were delivering between $1.5 billion and $2.0 billion in tangible annual business value—an increase from the earlier $1.0 billion to $1.5 billion estimate. This upward revision reflected growing confidence as initiatives matured.​

Specific 2024 impacts included:

• Asset & Wealth Management: Generative AI tools contributed to a 20% year-over-year increase in gross sales between 2023 and 2024. The Coach AI tool enabled advisors to find information 95% faster and increased sales by 20%​

• Trading optimization: Algorithmic trading win rates improved from 52% to over 60% following deployment of deep reinforcement learning AI​

• Credit decisioning: 40% improvement in accuracy and 30% reduction in defaults, while loan approval rates increased 20-30%​

• Fraud detection and prevention: Cited as a “particularly significant contributor” to business value, with real-time fraud detection achieving 40% improvement in accuracy​

Harvard Business School published a case study in 2024 examining JPMorgan’s AI tools for 5,000 high-net-worth private banking client advisors, documenting nearly $1.5 billion in cumulative savings through fraud prevention, personalization, trading efficiencies, operational improvements, and credit decisions.​

2025-2026: Consistent $2B Annual Value

In October 2025, CEO Jamie Dimon confirmed that AI was saving “about $2 billion of benefit” annually, stating this figure represented ongoing value being realized across the organization. This $2 billion in annual business value was being generated from approximately $2 billion in annual AI investment—a 1:1 return in direct measurable benefits, not accounting for harder-to-quantify productivity gains, competitive positioning, or future option value.​

Dimon characterized the $2 billion figure as “the tip of the iceberg,” suggesting that total AI value—including productivity gains, faster product development, improved client experiences, and talent attraction—was substantially higher than the measured financial returns.​

Cost Avoidance and Defensive Value

A critical but often overlooked component of AI ROI is defensive cost avoidance—preventing costs that would otherwise occur. JPMorgan faces a 12% compound annual growth rate in fraud attacks. Without AI-powered defenses, fraud losses would escalate exponentially. By using AI to hold fraud costs flat despite surging attack volumes, the bank avoids hundreds of millions in potential losses annually.​

The U.S. Treasury’s recovery of over $375 million through AI-powered check fraud detection—using technology influenced by JPMorgan’s approaches—illustrates the magnitude of fraud losses that advanced detection can prevent.​

IV. AI in Jamie Dimon’s Shareholder Letters: Strategic Evolution

Jamie Dimon’s annual letters to shareholders serve as the bank's authoritative articulation of its strategic priorities and long-term perspective. The evolution of AI discussion in these letters tracks the technology’s rising importance within the firm’s competitive strategy.

Thematic Consistency

Across the shareholder letters, several themes recur:

1. Long-term investment horizon: Dimon consistently frames technology investment as essential for positioning the company “to grow and prosper for decades”, explicitly rejecting short-term earnings optimization that sacrifices future competitiveness.​

2. Competitive imperative: The letters acknowledge that “the walls that protect this company are not particularly high — and we face extraordinary competition”. AI investment is positioned as defensive and offensive: necessary to avoid being disrupted while creating a competitive advantage.​

3. Financial discipline: Despite massive spending, Dimon emphasizes that technology investments “pay for themselves” and generate “strong returns”, rejecting the characterization of technology spending as unproductive overhead.​

4. Talent as foundation: The letters repeatedly emphasize that “the foundation of our success rests with our people”, with technology investments meaningless without the human capital to build, deploy, and manage AI systems.​

V. Earnings Call Evolution: From Vision to Accountability (2022-2026)

Quarterly earnings calls provide a different lens than annual shareholder letters: they reflect investor questions, near-term results, and management accountability for spending commitments. The evolution of AI discussion in these forums reveals increasing specificity as the technology moved from strategic vision to operational execution.

2022: Building Infrastructure, Limited Public Detail

The Q4 2022 earnings call (January 2023) focused heavily on macroeconomic outlook, credit reserve methodology, and interest rate impacts. AI was not prominently featured in the available transcript, suggesting it remained primarily an internal priority not yet central to the external investment narrative. The focus was on defending strong financial performance ($132.3 billion revenue, $37.7 billion net income, 18% ROTCE) amid economic uncertainty.​

2023: Investor Day Breakthrough and Measurement Commitments

The May 2023 Investor Day represented a breakthrough in public AI disclosure. CIO Lori Beer provided granular metrics: more than 300 use cases in production (up 34% year-over-year), $1 billion commitment raised to $1.5 billion in business value for 2023, AI ROI increasing 25% from 2021 to 2022, and detailed talent numbers (900+ data scientists, 600 ML engineers, 200-person research team).​

Beer also disclosed specific use-case returns: $220 million in benefits from retail personalization and credit card upgrades, and $100 million in benefits from commercial banking sales intelligence. This level of financial specificity was unprecedented among large banks and created a new standard for AI ROI transparency.​

The Investor Day also addressed generative AI directly. Beer stated: “We couldn’t discuss AI without mentioning GPT and large language models. We recognize the power and opportunity of these tools”. This acknowledgment—six months after ChatGPT’s launch—signaled JPMorgan was actively evaluating generative AI while continuing to scale proven predictive AI applications.​

2024: Disciplined Execution and Efficiency Focus

The Q1 2024 earnings call (January) featured CFO Jeremy Barnum articulating a “pragmatic and disciplined approach” to AI, emphasizing a “contained, well-chosen list of high-impact use cases” rather than “chasing shiny objects”. This framing responded to the generative AI hype cycle: Barnum wanted investors to understand JPMorgan was focused on results, not experimentation.​

Barnum noted the appointment of Teresa Heitsenrether as Chief Data and Analytics Officer as evidence of AI’s strategic priority. He emphasized: “We’re holding ourselves accountable for actual results”.​

By Q3 2024 (October), Barnum acknowledged that consensus expense forecasts for 2025 appeared “conservative” and highlighted that increases represented “investments in growth areas such as branches, wealth management, technology, and AI” expected to “yield positive returns”. This defensive posture suggested investor concern about rising expenses, with management positioning AI as a growth investment rather than an unproductive cost.​

2025: Efficiency Gains and Headcount Discipline

The May 2025 Investor Day featured CFO Barnum emphasizing the $18 billion technology budget (6% increase over 2024) and CEO Dimon’s memorable framing: “It’s table stakes. It will be for the rest of eternity”. This statement positioned ongoing technology investment as a permanent structural reality of banking, not a cyclical discretionary spend.​

Barnum revealed that approximately 65% of applications now run processing in the cloud (up from 50% prior year) and ~80% run on modern infrastructure (up from 75%). He also noted that while overall expenses increased $26 billion over the prior five years, revenue grew $54 billion over the same period—demonstrating positive operating leverage despite heavy investment.​

Critically, Barnum stated that the bank’s 317,000 employees represented inefficiencies in headcount introduced during rapid growth, and that management was now asking teams to “resist headcount growth where possible and increase their focus on efficiency”. AI-driven productivity gains gave management confidence in constraining headcount: “We’re going to do the same this year: have a very strong bias against having the reflective response to any given need to hire more people and feeling a little bit more confident in our ability to put that pressure on the organization because we know that, even if we can’t always measure it that precisely, there are definitely productivity tailwinds from AI”.​

2026: Competitive Imperative and Long-Term Defense

The Q4 2024 earnings call (January 2026) featured the most forceful public defense of technology spending in the five-year period. Facing questions from Wells Fargo analyst Mike Mayo about the $9.7 billion projected expense increase for 2026, Dimon responded:

“We’re not going to try to meet some expense target, and then 10 years from now, you’d be asking us a question, how did JPMorgan get left behind?”​

Dimon positioned the spending as competitive necessity: “We’re going to stay out front, so help us God”. He emphasized competition not just from other banks but from fintech firms like Affirm, SoFi, and Revolut, stating “they’re good players” and JPMorgan needed to outcompete them.​

On AI specifically, Dimon said it was “not a big driver” of the $9.7 billion increase but would “likely drive future efficiency”. This nuance was important: the 2026 expense increase was driven more by volume growth, inflation, and strategic expansion (branches, wealth management) than by AI spending itself. However, AI spending would continue, and its efficiency benefits would help offset other cost pressures.​

Dimon declined to provide specific AI investment figures, stating he had already been “quite blunt” but wouldn’t disclose details that could create “competitive disadvantage”. On anticipated returns, he said simply: “Part of it is to trust me, I’m sorry”. This represented a rare appeal to investor confidence rather than quantitative justification—perhaps reflecting that some AI investments involve long-term competitive positioning whose financial returns are difficult to project precisely.​

VI. Specific AI Applications and Measured Impacts

AI delivers heterogeneous but significant impacts across JPMorgan's operations: 95% faster information retrieval for advisors transforms client service economics, while 40% improvements in fraud and credit accuracy directly reduce losses. Even the "modest" 15.4 percentage point trading win rate improvement (52% to 60%) generates millions in daily P&L gains on $260B trading volume.

While aggregate financial returns demonstrate AI’s value, understanding specific use cases reveals how the technology creates value across diverse banking operations.

Fraud Detection and Risk Management

JPMorgan’s AI-powered fraud detection represents one of the highest-value applications, combining revenue protection (preventing losses) with cost avoidance (holding fraud costs flat despite surging attacks).

Real-time fraud detection systems analyze transaction patterns across massive datasets, distinguishing legitimate activity from fraudulent behavior with 40% better accuracy than traditional rule-based systems. The systems identify fraud patterns invisible to human analysts by processing correlations across millions of accounts simultaneously.​

The bank faces a 12% compound annual growth rate in fraud attacks. Without AI, fraud losses would grow exponentially. By using AI to detect and prevent fraud proactively, JPMorgan holds fraud costs flat—a massive cost avoidance equivalent to hundreds of millions in annual savings. Cumulative fraud prevention value from AI has been estimated at over $1.5 billion when combined with other security enhancements.​

The U.S. Treasury Department adopted AI-powered check fraud detection methodologies influenced by JPMorgan’s technology, recovering over $375 million. This third-party validation demonstrates the effectiveness of JPMorgan’s fraud AI at institutional scale.​

Algorithmic Trading and Markets

JPMorgan has deployed AI on trading desks for over eight years as of 2025, using machine learning to optimize execution, identify market signals, and manage portfolio construction. The bank trades approximately $260 billion in volume daily, reaching $500 billion during peak periods.​

Specific impacts include:

• Trading win rates: Deep reinforcement learning algorithms increased algorithmic trading win rates from 52% to over 60%—a substantial improvement that translates to millions in daily P&L gains​

• Optimal execution: AI systems identify the best timing, venue, and sizing for trades across multiple data streams (news feeds, social media sentiment, economic indicators, historical patterns) at machine speed​

• Pattern recognition: AI detects arbitrage opportunities and market signals that exist for fractions of a second, enabling the bank to capitalize on micro-opportunities inaccessible to human traders​

The adaptive nature of these systems is critical: machine learning models continuously refine trading logic based on changing market conditions without manual intervention, effectively creating a trader who “never sleeps, never gets emotional, and gets smarter with every transaction”.​

Wealth Management and Client Advisory

The Coach AI tool deployed for JPMorgan’s private client advisors represents a breakthrough in AI-augmented knowledge work. This generative AI application helps advisors retrieve research, answer client questions, and create personalized recommendations at dramatically faster speeds.

Measured impacts include:

• 95% faster information retrieval: Advisors can find relevant content and research 95% faster than manual searching, freeing time for client conversations rather than document hunting​

• 20% sales increase: Between 2023 and 2024, Asset & Wealth Management saw a 20% year-over-year increase in gross sales, with GenAI tools cited as a key contributor​

• 50% client base expansion potential: JPMorgan projects advisors will be able to expand their client bases by 50% over 3-5 years as AI handles research tasks, enabling each advisor to serve more clients without quality degradation​

During the April 2025 market volatility triggered by U.S. tariff announcements, Coach AI enabled advisors to rapidly provide personalized guidance to anxious clients. Mary Erdoes, CEO of Asset & Wealth Management, noted: “The market has experienced several fluctuations that are not typical, making it quite challenging to consider all your clients and their various needs. AI tools enabled advisors to swiftly address client inquiries by analyzing their trading behaviors and predicting their questions”.​

The tool also performs “anticipatory work,” helping advisors prepare for client conversations by pre-analyzing portfolios and generating scenario analyses. This predictive capability transforms advisor-client relationships from reactive to proactive.​

Credit Decisioning and Underwriting

AI-powered credit models have improved both the speed and accuracy of lending decisions while expanding access to previously underserved borrowers.

Documented improvements include:

• 40% accuracy improvement: AI models predict default risk with 40% better accuracy than traditional credit scoring​

• 30% reduction in defaults: Better risk assessment translates to lower actual default rates​

• 20-30% increase in approval rates: By incorporating non-traditional data sources (cash flow patterns, transaction behaviors, utility payments), AI extends credit to applicants who would have been declined under traditional FICO-based models​

Speed gains are equally important: loan decisions that previously took hours or days now occur in seconds, enabling real-time lending decisions at the point of sale. This speed advantage is particularly valuable in competitive consumer lending markets where applicants often apply to multiple lenders simultaneously.​

The inclusiveness dimension is noteworthy: by assessing creditworthiness through behavioral data rather than just traditional credit history, AI enables JPMorgan to serve populations historically excluded from banking (thin-file borrowers, immigrants, young adults without established credit).​

Operations and Back-Office Automation

Operations represents one of the largest opportunity areas for AI-driven efficiency, and also the area where workforce displacement is most visible.

Productivity gains: Operations productivity historically improved at about 3% annually but has accelerated to approximately 6% annual improvement with AI deployment. JPMorgan expects each operations specialist to become 40-50% more productive over time as AI automates routine tasks.​

Agentic AI for complex workflows: The bank is deploying autonomous AI agents that execute multi-step processes independently. Examples include:

• Investment bankers creating 5-page presentation decks in 30 seconds (work that previously took junior analysts hours)​

• Lawyers scanning and generating contracts with covenant extraction occurring instantly​

• Trade settlement and account setup processes handled end-to-end by AI agents​

Call center enhancement: The EVEE Intelligent Q&A tool deployed in call centers delivers context-aware responses, improving resolution times and customer satisfaction.​

Workforce implications: Consumer banking chief Marianne Lake announced in 2025 that operations staff would decline at least 10% as agentic AI takes over complex tasks. This represents one of the first explicit headcount-reduction commitments by a major bank CEO, directly attributed to AI.​

New roles are emerging to support AI operations: “context engineers” ensuring AI systems have proper information, knowledge management specialists, and up-skilled software engineers building agentic systems. However, the net employment effect is clearly negative for traditional operations roles.​

Software Engineering and Code Generation

AI is transforming JPMorgan’s software development process, a critical capability given the bank’s 60,000+ technology employees and $18 billion technology budget.

Code generation at scale: Alphabet CEO Sundar Pichai stated in 2024 that “nearly half of all code [is] generated by AI”. JPMorgan has implemented similar capabilities. Wells Fargo reported in Q3 2025 that there had been “more than 1 million AI-driven automated code reviews this year, creating 100,000 hours of weekly capacity for developers”—providing a peer benchmark for JPMorgan’s likely gains.​

Code deployment velocity: JPMorgan has increased code deployments by over 70% over the past two years while achieving a 20% reduction in work being replanned—demonstrating that AI-assisted development is both faster and higher quality.​

Investment banking automation: Investment bankers are automating 40% of research tasks using AI tools, and portfolio managers have cut research time by as much as 83%. These gains enable senior professionals to focus on client relationships and strategic advisory rather than information gathering.​

Subscribe now

VII. Strategic Implications and Competitive Positioning

JPMorgan’s AI strategy has profound implications for global banking's competitive dynamics. The bank’s scale, investment capacity, and execution discipline are widening the gap with smaller competitors.

The AI Moat: Capital, Talent, and Data

Three mutually reinforcing advantages position JPMorgan to sustain AI leadership:

1. Financial capacity: With $180.6 billion in annual revenue and $58.5 billion in net income, JPMorgan can sustain $18 billion annual technology budgets and $2 billion annual AI spending that smaller banks cannot match. Regional banks with $5-10 billion in annual revenue cannot allocate equivalent percentages to technology without destroying profitability.​

2. Talent concentration: JPMorgan employs over 2,000 AI/ML experts and data scientists—more than the total technology workforce of many regional banks. The bank offers base salaries exceeding $275,000 for AI specialists, competing directly with Big Tech for talent. Smaller institutions cannot match this compensation or offer the scale and complexity of problems that attract top AI researchers.​

3. Data volume and velocity: JPMorgan processes nearly $10 trillion in payments daily across 120 currencies and more than 160 countries, safeguarding over $32 trillion in assets. This “data flywheel” provides an unparalleled source of proprietary training data for AI models. A regional bank processing $50 billion in daily payments—1/200th JPMorgan’s volume—simply cannot train models with equivalent predictive power.​

Daniel Pinto, JPMorgan’s President and COO, characterized the firm’s advantage as a “data flywheel”: “a virtuous cycle powered by over $10 trillion in daily transactions, which provides an unparalleled source of high-stakes, proprietary data for training superior AI models”.​

Valuation Premium and Market Recognition

JPMorgan’s AI execution has translated into a persistent valuation premium versus banking peers. The stock trades at approximately 2.68x book value and 15-16x forward earnings—multiples that exceed large-cap bank averages and reflect investor confidence in sustained competitive advantage.​

Analyst commentary explicitly links this premium to AI productivity:

• “JPMorgan trades at a premium as AI productivity offsets rising cost guidance”​

• “The combination of earnings power, AI-driven productivity, non-interest income growth and robust credit metrics means JPM still justifies a Buy-tilted stance”​

• Wells Fargo analyst Mike Mayo: “AI financial gains could rise by an additional billion dollars for the bank”​

The stock advanced approximately 35% in 2025, outperforming bank indices and justifying the premium through demonstrated profitability. Investors are effectively paying for the expectation that JPMorgan’s AI-driven efficiency and margin expansion will persist for years, creating a structural advantage over peers.​

Competitive Response and Industry Dynamics

Other large banks are responding to JPMorgan’s AI lead, but gaps remain substantial:

Wells Fargo: Mentioned AI only once during its Q3 2025 earnings call, highlighting automation efforts to drive efficiency. CFO Michael Santomassimo noted: “AI helps on some of that for sure,” but provided no specific use cases, financial returns, or strategic roadmap. This vague treatment suggests Wells Fargo is in early AI deployment stages, years behind JPMorgan.​

Citigroup: Co-CIO Shadam Zafar outlined a “four phases” AI strategy with immediate focus on productivity and back-office use cases. Citigroup has approximately 30,000 employees using AI for software engineering and platform modernization. However, the bank has not disclosed specific financial returns or use case counts, suggesting it is still building foundational capabilities.​

Goldman Sachs: Introducing generative AI assistants for bankers, traders, and asset managers, but has not disclosed production use case counts or financial returns. CEO David Solomon stated on Q3 2025 call: “We are drilling in on a handful of front-to-back work streams that can significantly benefit from AI-driven process reengineering”—suggesting early-stage focus on identifying opportunities rather than scaled deployment.​

Bank of America: Has implemented AI in call centers and operational processes, but public disclosures lack the granularity of JPMorgan’s reporting.​

A Consistent Pattern: competitors acknowledge AI’s importance and are investing, but none have disclosed production use case counts, specific financial returns, or organizational scale approaching JPMorgan’s transparency and demonstrated execution.

Fintech Competition and Defensive Positioning

Jamie Dimon’s January 2026 statement that JPMorgan competes not just with banks but with fintech firms like Affirm, SoFi, and Revolut reflects a critical strategic reality. These fintech challengers are digital-native, operate on modern technology stacks, and lack legacy infrastructure constraints.​

JPMorgan’s AI strategy is explicitly defensive against fintech disruption:

• User experience parity: AI-powered personalization (credit card upgrades, product recommendations, real-time fraud alerts) matches or exceeds fintech UX while leveraging JPMorgan’s broader product suite​

• Speed and automation: Real-time credit decisions, instant account setup, and 24/7 AI-powered service match fintech convenience expectations​

• Cost efficiency: AI-driven operations productivity gains enable JPMorgan to compete on cost despite legacy infrastructure burdens​

Dimon’s comment—”We’re going to stay out front, so help us God”—reveals leadership concern that complacency could allow nimbler fintech competitors to capture younger customer cohorts before they develop banking relationships with incumbents.​

Subscribe now

VIII. Workforce Transformation and Human Capital Implications

JPMorgan’s AI deployment is visibly reshaping workforce composition, a transformation the bank has addressed with unusual candor compared to peers emphasizing only positive “augmentation” narratives.

Explicit Headcount Discipline

In May 2025, CFO Jeremy Barnum stated that JPMorgan’s 317,000 employees represented some inefficiency introduced during rapid growth, and management was asking teams to “resist headcount growth where possible”. This marked a strategic shift: rather than allowing AI to create capacity for more hiring, the bank would capture AI productivity gains as margin expansion.​

Barnum explained the logic: “We’re going to have a very strong bias against having the reflective response to any given need to hire more people and feeling a little bit more confident in our ability to put that pressure on the organization because we know that, even if we can’t always measure it that precisely, there are definitely productivity tailwinds from AI”.​

In 2025, Consumer Banking Chief Marianne Lake announced that operations staff would decline at least 10% as agentic AI takes over complex tasks. This represents one of the first explicit headcount reduction targets directly attributed to AI by a major U.S. bank CEO. The 10% figure applied specifically to operations—back-office roles handling account setup, fraud investigation, trade settlement, and similar tasks—rather than firmwide employment.​

Differential Impact by Role Type

AI’s workforce impact is highly non-uniform:

Beneficiaries (augmented roles):

• Private bankers and wealth advisors: AI handles research, enabling advisors to serve 50% more clients​

• Traders and investment bankers: AI automates 40% of research tasks, freeing time for client relationships and deal structuring​

• Software developers: AI-powered code generation increases deployment velocity 70% while reducing rework 20%​

• Risk managers: AI identifies fraud patterns and risk signals invisible to human analysis​

At-risk roles (displaced or significantly reduced):

• Operations specialists: Back-office processing automated by agentic AI, leading to 10% headcount reduction​

• Junior analysts: Research and presentation creation automated, reducing entry-level hiring needs​

• Call center agents: EVEE AI handles routine inquiries, reducing human escalation needs​

• Administrative staff: Document processing, contract review, and compliance tasks automated​

Emerging roles (new categories):

• Context engineers: Ensure AI systems have proper information access and retrieval​

• Knowledge management specialists: Curate and structure data for AI consumption​

• AI trainers and validators: Ensure model outputs align with regulatory and quality standards

• Agentic AI supervisors: Monitor and manage autonomous AI agent workflows

Generational and Entry-Level Implications

Stanford researchers analyzing ADP data found that early-career workers (ages 22-25) in AI-exposed occupations experienced a 6% employment decline from late 2022 to July 2025. This suggests AI is reducing entry-level hiring across the economy, not just in banking.​

For JPMorgan, this creates a management challenge: traditional career progression assumes large cohorts of junior analysts and associates who gain experience through repetitive tasks before advancing to senior roles. If AI eliminates 40% of junior analyst work, how does the bank develop future managing directors?

One answer, suggested by JPMorgan executives, is that junior employees will gain early management experience by overseeing AI agents—effectively managing “digital workers” rather than human teams. Whether this substitutes adequately for traditional skill development remains uncertain.​

Training and Change Management

JPMorgan has implemented extensive training programs to prepare employees for AI integration. The bank has trained “tens of thousands of employees” on using AI tools effectively in daily work. The LLM Suite platform has 200,000 users (63% of workforce), with over half using it multiple times daily.​

This democratization strategy—making AI accessible to broad employee populations rather than concentrating it in specialized teams—aims to capture productivity gains across the organization while building a culture of AI literacy.

X. Conclusion: The AI Transformation Imperative

JPMorgan Chase’s AI strategy from 2021 through 2026 represents the most comprehensive, transparent, and financially disciplined enterprise AI deployment in global banking. The bank has moved from 300 use cases and $100 million in annual value (2022) to 450+ use cases and $2 billion in annual value (2025), with a clear path to 1,000 use cases by 2026.

This transformation rests on three interdependent pillars:

1. Capital and Investment: $18 billion annual technology budgets, $2 billion dedicated AI spending, $2+ billion in cloud data center infrastructure, and willingness to absorb near-term expense increases for long-term competitive positioning.

2. Talent and Organization: Over 2,000 AI/ML experts and data scientists, 200-person research team, C-suite leadership with Chief Data and Analytics Officer on the Operating Committee, and firmwide AI democratization with 200,000 employees using AI tools daily.

3. Data and Scale: $10 trillion in daily transaction data, 84 million U.S. customers, 90% of Fortune 500 relationships, and proprietary datasets that create a self-reinforcing data flywheel impossible for smaller competitors to replicate.

The measured financial returns, $2 billion annually from $2 billion in AI investment, represent a 1:1 direct ROI, not accounting for harder-to-quantify benefits (faster product development, improved client experiences, talent attraction, competitive positioning, option value). CEO Dimon’s characterization of this as “the tip of the iceberg” suggests management believes total value substantially exceeds measured returns.​

JPMorgan’s AI leadership is visible in comparative metrics: the bank ranks #1 on Evident AI’s Index for AI maturity, employs more AI talent than any global peer, has disclosed more production use cases and specific financial returns than competitors combined, and commands a valuation premium (2.68x book, 15-16x forward earnings) that reflects investor confidence in sustained competitive advantage.​

The strategic necessity underlying this investment is captured in Jamie Dimon’s January 2026 statement: “We’re not going to try to meet some expense target, and then 10 years from now, you’d be asking us a question, how did JPMorgan get left behind?”. This reflects recognition that AI is not a discretionary technology investment but an existential competitive imperative—comparable to the introduction of ATMs, online banking, or mobile apps in previous eras.​

The workforce transformation accompanying AI deployment is profound and irreversible. Operations staff declining 10%, junior analyst tasks automated 40%, operations productivity doubling from 3% to 6% annual improvement, and headcount discipline despite revenue growth all signal a structural shift in banking’s labor economics. JPMorgan’s candor about these changes—rare among large employers—reflects confidence that efficiency gains will outweigh the organizational friction of workforce reduction.​

Looking forward, JPMorgan’s AI strategy positions the bank to sustain competitive advantage through the next decade. The combination of scale, data, talent, capital, and execution discipline creates a widening moat versus smaller competitors who lack the resources for equivalent investment. Regional banks spending $200-300 million annually on technology cannot match JPMorgan’s $18 billion budget; community banks with 50 data scientists cannot compete with 2,000+ AI specialists.

However, risks remain material. Execution challenges in scaling from 450 to 1,000 use cases, potential commoditization of AI capabilities through cloud APIs, intensifying cybersecurity threats, regulatory scrutiny of AI fairness and explainability, and potential credit cycle downturns all threaten the investment thesis.

The ultimate question is whether AI creates sustainable competitive advantage or evolves into table-stakes commodity functionality, similar to how ATMs and online banking transitioned from differentiators to universal expectations. JPMorgan is betting $2+ billion annually that early leadership, proprietary data, and execution excellence will create a persistent moat. The financial results from 2021-2026 suggest that bet is paying off, but the full return will only be visible over the coming decade.

For business strategists, operators and economists studying the future of companies, JPMorgan Chase provides the most detailed case study available on how a $4 trillion financial institution navigates the AI transformation: with massive capital commitment, relentless focus on measured returns, candor about workforce displacement, and recognition that technology investment is no longer discretionary but existential.

References: Here is a Google Doc I put together with comprehensive links and access to all of the underlying research and data for the article.

I’m building something different in 2026: Personally working 1:1 with just 100 serious operators, founders, creators, and builders.

4 quarterly 60-minute working sessions with me (recorded for you) to build your operating strategy, crush team issues, sharpen your brand, or fix whatever’s blocking growth.

For the first 25 who join, I’ve slashed the Operating Founder tier from $550/year to $99/year. Once those spots fill, the price jumps back.

Why $99? I want to partner closely with 100 inventive builders in our AI-first era, learn your stories, and see what’s really working.

Subscribe now

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.

Get more from John Brewton in the Substack app

Available for iOS and Android

Get the app

Read more

In 2025, AI‑assisted tooling will let a single operator prototype most department‑level data pipelines at low cost in 8–12 weeks, provided they enforce basic governance and testing. This does not replace enterprise systems. It unlocks the unapproved projects in your backlog and converts domain experts into credible builders.

Read more