Banking is entering its biggest transformation since the arrival of the internet - and the winners may not be traditional banks.
I’m Ashish Chopra, a CIO/CTO, banking strategist, entrepreneur, and board advisor with 20+ years of experience helping regulated financial institutions modernize.
Here, I’ll share direct perspectives on:
→ AI and digital banking
→ Fintech and stablecoins
→ Technology leadership
→ What major announcements really mean
→ Questions every CEO, CIO, and board should be asking
Expect practical insights, honest opinions, and a few predictions worth debating.
If you’re building the future of financial services, follow along.
Jamie Dimon Saw the Cracks Before 2008. Then He Turned Crisis Into Opportunity.
While banks struggled, JPMorgan made two defining moves:
🏦 Bear Stearns: Acquired in a Fed-supported deal that expanded its Wall Street presence.
🏦 Washington Mutual: Acquired banking operations for $1.9B, adding ~2,200 branches and $188B in deposits.
By the end of 2008, @jpmorgan 's assets had grown from $1.56T to $2.18T.
The leadership lesson? Spotting risk isn't enough. The advantage comes from positioning early and acting when others can't.
Now JPMorgan is investing billions in AI, blockchain, and financial infrastructure.
Could 2027–2032 create an even bigger competitive divide than 2008?
I explore this shift in my forthcoming book, The Autonomous Economy.
#JamieDimon #JPMorgan #Banking #AI #LeadershipDevelopment
The stress-test point is fair. The replacements are still adding risk on top of the old system. For banks the practical reply is not to wait for perfect rails. It is to put the control plane — mandate, limit, revoke, named owner — in place before the agent volume arrives. Otherwise the new risk lands on the same balance sheet.
How much autonomy an agent should have over money is the question every bank is now answering in practice. Most still keep the human on the final decision. The institutions that lead will not be the ones with the most capable model. They will be the ones that can show identity, permission, audit, and a clear owner for every kill switch when the agent does act.
Token costs at 20 to 25 percent and human oversight at 70 to 75 percent is the number that reframes the agent. The human in the loop is not only a safeguard. It is currently the cost. BNY giving agents login IDs, employee numbers, and performance reviews under about 100 human managers is the organizational version of the same point. The agent can act. The human still owns the authority and the exception. That is the architecture that scales.
Ease of use is still the adoption gate. 63 percent through the current bank, 54 percent blocked by the ERP or treasury connection. That is not a technology problem. It is a distribution problem. The product that wins is the one that appears inside the system the controller already trusts, with the same mandate, limit, and revoke the bank already runs.
EY-Parthenon just measured the adoption path. 63 percent of corporates would take stablecoins or tokenized deposits through their current bank or payments provider.
54 percent say the barrier is that it does not connect to the ERP or treasury system they already use. They are not asking for a new rail. They are asking for the outcome inside the tools the finance team opens every morning.
The bank that embeds the deposit token and the agent mandate into the existing treasury workflow keeps the relationship. The one that asks the treasurer to log into a separate wallet loses it.
Banks are not fighting digital money. They are fighting someone else’s digital money. Three consortia, 52 percent of the top 50 US banks now talking tokenized deposits, a shared deposit network aimed at the first half of 2027. The open risk is the running cost and the control plane. Owning the token does not automatically mean owning the mandate when an agent initiates the payment.
Jackie Reses is right that the technology is not the hard part. Who is in charge, what the agent can do, when access ends, and how the data stays secure are standards problems. Banks and fintechs that wait for someone else to write those standards will inherit someone else’s liability. The ones that publish a clear permission model first will set the floor.
24/7 interbank settlement of tokenized deposits with tokenized central-bank money is the infrastructure step that matters. The control question arrives with it. When an agent can instruct a payment outside banking hours, the bank still needs a mandate, a limit, and a revoke that works at 2 a.m. Hong Kong is building both the rail and the supervision conversation at the same time. That is the right sequence.
Haseeb just named the incentive. Banks are not only blocking outside AI agents because the agents are unsafe. They are blocking them because an agent that can move cash to a higher yield is an agentic bank run on cheap retail deposits. JPM and Schwab already did it. The response that keeps the customer is not a permanent ban. It is a verified agent the bank can identify, limit, and shut down — and a deposit rate that does not rely on the customer never noticing.
SoFi offering both the bank-issued stablecoin and the tokenized deposit is the interoperability most institutions are still designing. The useful version lets a customer move between them 24/7 without the deposit leaving the bank’s books or the insurance perimeter. That is the product. The chain is just the rail.
63 percent of financial institutions planning tokenized asset settlement on blockchain rails is the adoption number that matters. Collaboration, not competition, is the right tone. The practical test is whether the settlement still carries the bank’s mandate, limit, and revoke when the asset and the cash leg both move onchain.
Lose the lunch or participate without cutting into lending profits is the real choice. Stablecoins that pull transactional balances are a funding problem. Tokenized deposits that still count as cash and still support credit lines are the participation path. The bank that offers the second keeps the margin. The one that only bans the first loses the customer.
Freeze and clawback as the reason for the chain choice is the clearest bank statement in weeks. Putting regulated money on a public blockchain only works if those controls are native and the uptime is bank-grade. The open question is whether the same freeze and revoke still fire when an agent, not a human, initiates the payment.
Jamie Walker at U.S. Bank just said why they chose Stellar. Asset freezing and clawback are table stakes when you put regulated money on a public chain.
Resilience and trust are the rest. If the network is down for a minute, confidence is gone. That is the bank position in one sentence.
Tokenized deposits only work if the institution can still freeze, claw, and prove the system stayed up. Speed without those controls is just a faster way to lose the liability.
A UK bank with roughly £7 billion in deposits targeting £250 million in tokenized deposits is past the concept stage. Midnight as the privacy layer only matters if the deposit remains a claim on the bank and the controls — freeze, limit, revoke — stay with the institution. Next week’s updates will show whether this is a pilot or a product.
Stack, not fight, is the right map. Deposit tokens as reserves inside the bank perimeter, paying interest. Stablecoins as bearer money anyone can hold. An issuer that can rebalance between them in code at 3 a.m. on a Sunday is offering a better experience. The bank still has to own which reserve sits under which liability.
Stablecoins as a deposit-growth path is the framing most banks still miss. Idle balances from stablecoin receipts moving into tokenized deposits that count as cash and support credit lines is the business case. The early movers keep both the payment and the liability. The control that has to travel is the mandate and the revoke, or the deposit is just a faster way to leave the book.
Stopping hacker bots on stolen logins is correct. Locking out the customer’s own agent is not a strategy. A key the customer can hand over, with limits they set and the bank can still revoke, is the product. A permanent ban just moves the relationship to whoever will host the agent.
@RBC just blocked customers’ personal AI agents from online banking.
The worry is real — stolen logins and rogue bots. The response is not a permanent lock.
Give the customer a key they can hand their own agent, with limits they set and a revoke the bank still controls. Otherwise the cash moves to the institution that will host the agent.
The same week @tempo noted the other side: businesses paid in stablecoins will park idle balances in tokenized deposits that still count as cash and support credit.
Banks that offer the governed version early win both the agent and the deposit.