This is an unusually clear case study of what “AI transformation” actually looks like when it’s treated like infrastructure + governance + measurement, not a demo! What I appreciate most is the clinical-like discipline JPM brings: define the problem, build the substrate (data + platforms), standardize workflows, and then hold the system accountable to outcomes. The “we’re holding ourselves accountable for actual results” mindset (and the explicit $100M → $2B value narrative) is exactly the kind of measurement culture most organizations lack, whether in banking or healthcare. 
Also, the sequencing matters: this reads less like “AI is magic” and more like “AI is a force multiplier once you’ve done the unglamorous work”; cloud migration, modern infrastructure, org design, and incentives aligned to productivity rather than headcount reflexes. 
As a physician-scientist, I couldn’t help seeing the parallel to medicine: we keep trying to bolt “AI” onto fragmented data and variable workflows, then act surprised when impact is modest. The real lesson here is boring, but decisive: data quality, standardization, and accountability are the intervention; the model is just the drug delivery system.
What makes this compelling isn’t the $2B number, but the sequencing. A decade of data unification before visible AI returns is a reminder that most “fast AI transformations” are skipping the hard, unglamorous prerequisites. Excellent analysis!
A rigorous, operator-level case study that cuts through the AI hype and shows how disciplined execution, not experimentation theater, turns sustained investment into measurable advantage
Thanks for bringing up that distinction! Most organizations are stuck in the theater of pilots, while JPMorgan has spent a decade building an industrial production line for models.
Can't deny the data! You can trace the line directly from their infrastructure spending in 2021 to the efficiency gains showing up in the 2024 earnings.
The biography of JP Morgan is a great read. Gets less compelling as you move away from JP and get to the grandson but heyho. After reading this I think it needs an update as recently they've been of fire
It makes complete sense to pause and build from intention rather than just chase output, and you’ve clearly set yourself up for direction rather than distraction this year.
As you experiment further, what part of your new routine do you think will have the biggest impact by the end of February?
Yet another unbelievable analysis John. I can't think of anyone writing on Substack right now who goes into as much depth and rigour as you do on these posts. Really interested to know: do you think there are any other companies that have adopted JPMorgan Chase's approach for long-term strategic thinking in regards to AI who might be about to reap the dividends in the next six months or so?
If we’re sticking to banking, Capital One is the most likely peer. They were cloud-native years before the rest of the industry and have a similar engineering first culture that is finally hitting its stride.
Outside of finance, I love what Walmart is doing with Data Ventures and their AI-native supply chain tech that moved from experimentation to massive operational dividends just last year.
Yes, I think I do in OUR world. We're around it everyday (you and I ) here in social media land... and I see it in the news with huge companies, but I'm not sure it is being fully adopted by small-mid caps... how about you?
You feed it, it grows, such a crazy idea… What happens when everything and everyone is maximized? What then? Are we talking about the right stuff, or just fluff? Not trying to be negative Nancy here but damn, this is a lot of excitement for something that will be ubiquitous. What comes after redundancy?
Agreed about the inevitable ubiquity and subsequent commoditization.
But that doesn’t means a valuable arbitrage doesn’t exist in the space between now and when that’s the case. Many winners and losers to be picked between now and then. Moreover, if ubiquity is inevitable, then isn’t there great value in understanding what the best are doing at the outset and how they’re implementing?
I believe human fallibility compounds the further we get from intrinsically created value. Our brains develop nuance through experiences of life as it is lived. If you no longer need to get on a bike in order to learn to ride, do you really know how to ride a bike? Probably not the best example, but it’s still early.
For two years, the corporate narrative has been 'Co-pilot/Augmentation' to avoid spooking the herd. JPM is the first major player to say the quiet part out loud:
The ROI of AI comes from Non-Linear Revenue Growth per Employee.
If you aren't severing the link between 'Growth' and 'Headcount,' you aren't doing an AI transformation. You're just adding software costs.
Thanks for this! Decoupling headcount from revenue growth is the only way to achieve true AI transformation. Everything else is just incremental software improvement.
This is an unusually clear case study of what “AI transformation” actually looks like when it’s treated like infrastructure + governance + measurement, not a demo! What I appreciate most is the clinical-like discipline JPM brings: define the problem, build the substrate (data + platforms), standardize workflows, and then hold the system accountable to outcomes. The “we’re holding ourselves accountable for actual results” mindset (and the explicit $100M → $2B value narrative) is exactly the kind of measurement culture most organizations lack, whether in banking or healthcare. 
Also, the sequencing matters: this reads less like “AI is magic” and more like “AI is a force multiplier once you’ve done the unglamorous work”; cloud migration, modern infrastructure, org design, and incentives aligned to productivity rather than headcount reflexes. 
As a physician-scientist, I couldn’t help seeing the parallel to medicine: we keep trying to bolt “AI” onto fragmented data and variable workflows, then act surprised when impact is modest. The real lesson here is boring, but decisive: data quality, standardization, and accountability are the intervention; the model is just the drug delivery system.
which operating habit do you think smaller organisations most underestimate when they look at institutions like this?
What makes this compelling isn’t the $2B number, but the sequencing. A decade of data unification before visible AI returns is a reminder that most “fast AI transformations” are skipping the hard, unglamorous prerequisites. Excellent analysis!
Thank you!
A rigorous, operator-level case study that cuts through the AI hype and shows how disciplined execution, not experimentation theater, turns sustained investment into measurable advantage
Thanks for bringing up that distinction! Most organizations are stuck in the theater of pilots, while JPMorgan has spent a decade building an industrial production line for models.
JPMorgan’s EPS trajectory from 2021–2024 highlights a clear journey
Can't deny the data! You can trace the line directly from their infrastructure spending in 2021 to the efficiency gains showing up in the 2024 earnings.
Glad you pointed that out, Daniel, the arc matters more than any single quarter.
The biography of JP Morgan is a great read. Gets less compelling as you move away from JP and get to the grandson but heyho. After reading this I think it needs an update as recently they've been of fire
Chris, please 🤲🏿😭 I already have too many books on my nightstand!
Appreciate the note, Chris, the last few years have definitely made the story more interesting again.
It makes complete sense to pause and build from intention rather than just chase output, and you’ve clearly set yourself up for direction rather than distraction this year.
As you experiment further, what part of your new routine do you think will have the biggest impact by the end of February?
Thank you, Melanie, chasing output without intention always catches up eventually.
Thank you, Melanie!
This is incredible, thanks for this. When you build a framework around implementation, there's huge gains to be seen.
Implementation always determines the outcome. A great strategy without the underlying operational plumbing is just a high cost research project.
Well put :=)
Appreciate this, Dennis, the idea is easy, but the doing is what makes it real.
Isn't that a truth of life in general 😀
Yet another unbelievable analysis John. I can't think of anyone writing on Substack right now who goes into as much depth and rigour as you do on these posts. Really interested to know: do you think there are any other companies that have adopted JPMorgan Chase's approach for long-term strategic thinking in regards to AI who might be about to reap the dividends in the next six months or so?
Sam asks the trillion dollar question!
If we’re sticking to banking, Capital One is the most likely peer. They were cloud-native years before the rest of the industry and have a similar engineering first culture that is finally hitting its stride.
Outside of finance, I love what Walmart is doing with Data Ventures and their AI-native supply chain tech that moved from experimentation to massive operational dividends just last year.
I appreciate you, Sam, I have found most firms talk more than they build right now.
Great analysis. The shift from viewing AI as a project to an existential framework is the key takeaway here. 🔥
Thank you, Dennis, do you see many leaders actually treating it that way yet?
Yes, I think I do in OUR world. We're around it everyday (you and I ) here in social media land... and I see it in the news with huge companies, but I'm not sure it is being fully adopted by small-mid caps... how about you?
Yes! When you stop treating AI as an add-on and start treating it as the core operating system, the economics of the entire company change.
wow impressive research and comment thanks !
You feed it, it grows, such a crazy idea… What happens when everything and everyone is maximized? What then? Are we talking about the right stuff, or just fluff? Not trying to be negative Nancy here but damn, this is a lot of excitement for something that will be ubiquitous. What comes after redundancy?
Great perspective!
Agreed about the inevitable ubiquity and subsequent commoditization.
But that doesn’t means a valuable arbitrage doesn’t exist in the space between now and when that’s the case. Many winners and losers to be picked between now and then. Moreover, if ubiquity is inevitable, then isn’t there great value in understanding what the best are doing at the outset and how they’re implementing?
I believe human fallibility compounds the further we get from intrinsically created value. Our brains develop nuance through experiences of life as it is lived. If you no longer need to get on a bike in order to learn to ride, do you really know how to ride a bike? Probably not the best example, but it’s still early.
For two years, the corporate narrative has been 'Co-pilot/Augmentation' to avoid spooking the herd. JPM is the first major player to say the quiet part out loud:
The ROI of AI comes from Non-Linear Revenue Growth per Employee.
If you aren't severing the link between 'Growth' and 'Headcount,' you aren't doing an AI transformation. You're just adding software costs.
Thanks for this! Decoupling headcount from revenue growth is the only way to achieve true AI transformation. Everything else is just incremental software improvement.
Thanks! POC hell is usually a symptom of poor data and engineering foundations. JPMorgan solved that problem years before ChatGPT even launched.