AI banking assistants could help customers compare providers, but safe account access alone does not prove fair advice. X Money, Revolut and banks would need to show alternatives, net benefits, incentives, permission limits and who fixes mistakes.
Would your bank’s AI tell you when a competing account is better for your money?
Schwab is rolling out its own assistant and planning controlled access for outside AI. JPMorgan is blocking some agents that use shared login credentials. Those controls address a real risk. They also leave X and Revolut a strategic question: can they earn trust by helping customers choose across providers?
A tool that can move money should not inherit every power that comes with your password. Once it has safe, limited access, however, the customer still needs to know whose interests guide its recommendations.
X and Revolut have an opportunity here. They could compete for the role of helping people choose among financial providers, even when those people keep accounts elsewhere. To earn that role, they would need to show when a competing product serves the customer better—including when the customer should leave X or Revolut.
What Schwab is actually restricting
Calling this a blanket ban on outside technology misses an important distinction. Schwab’s data-access documents describe approved connections that let customers share selected account information without handing over their passwords. Customers can review and withdraw access. Those arrangements concern data access; they do not grant every connected tool permission to trade or move money.
In an interview transcript published on Schwab’s newsroom, CEO Rick Wurster describes two paths: Schwab’s own Charley assistant, and outside AI connected through a controlled system. The latter is a stated direction, not proof that every customer can use every outside model today.
Charley’s September announcement describes a phased rollout to eligible U.S. retail clients. Its listed launch actions include managing watchlists, alerts and certain account details. That is narrower than an assistant freely managing the customer’s entire financial life.
Schwab is trying to shape the route through which AI reaches the account. That can protect customers and preserve control of the experience at the same time. Neither fact establishes that its recommendations are biased.
A locked account does not settle the competition
Suppose an outside assistant cannot operate your brokerage account. It may still help you understand publicly available fees, compare products or prepare questions for the provider. Where permitted, you could supply selected account information yourself. A useful comparison need not begin with permission to move money.
The customer could then approve the next step through the institution’s own system. That is less seamless than full automation, but it leaves room for a competitor to become the place where the decision starts. It also avoids pretending that a data connection is permission to transact.
This broad shift is not a new discovery. McKinsey has described the possibility of AI becoming the main interface between customers and banks. A June economics preprint also examines how users could delegate account actions under verifiable controls. Neither establishes an unrestricted legal right for any agent to enter any account.
Compete on two promises together: access the customer can limit or withdraw, and a visible record of why the recommendation serves that customer.
A fintech could show the alternatives considered, the costs and conditions that changed the answer, and any financial benefit it receives. It should record when its own product loses. That would make the promise of acting for the customer easier to examine.
Revolut has a starting point—and a gap to close
Revolut already markets linked-account services in the UK. Its page advertises connections across 150 banks and 21 countries. That is company-stated coverage, not a promise that every account supports every action or that the same service is available in the United States.
It also began rolling out AIR, its conversational assistant, to its UK customer base in April. The launch describes spending insights, card controls and other tasks, with a fingerprint or face check to approve sensitive actions.
But its AI notice defines information rather than financial advice and limits the assessment of a customer’s situation beyond Revolut app activity. Linked accounts and AIR are therefore ingredients; their existence does not establish a complete assistant comparing and managing money across providers.
Revolut could build a clearer comparison service on top of approved connections. Start with a narrow job: identify spare cash after bills, compare eligible destinations after fees and access needs, and explain when keeping the money where it is makes sense. What the service actually does, and the rules it follows, would need to match that promise.
It could then publish repeatable examples where another provider wins, alongside examples where Revolut wins. A broad catalogue is useful. Evidence that the selection process follows the customer’s needs would be more persuasive.
X can start before the customer changes banks
X has a different entry point: conversations about money already occur on the platform. Its U.S. Money service is rolling out to selected users, and Cross River provides banking infrastructure behind the experience. X Payments is not an FDIC-insured bank.
X could use that setting to make financial comparisons easier to inspect. A reader could ask about a claim, see current provider terms, adjust the relevant assumptions and obtain a clear explanation of the trade-offs. Any access to accounts or execution of payments would remain a separate, explicitly authorized step.
That is a proposed strategy, not a claim that Grok currently manages accounts across banks. X’s published product information does not establish that service.
The strongest demonstration would include cases where X Money loses: the customer is ineligible, a paid plan costs more than the benefit, another account fits the need better, or moving bill money creates a problem. X’s own disclosures already show why conditions matter: rates and eligibility vary, fees can affect earnings, and New York residents do not receive interest on the stored-value accounts.
For Elon Musk, the strategic choice is straightforward to state and difficult to build: could X become the place people trust to make a financial decision, even when the resulting transaction happens elsewhere?
That approach could lead to a paid comparison service or other disclosed revenue. But it would have costs: current data, reliable calculations, compliance, support and conflict management. Recommending a competitor would not automatically create a viable business. The hypothesis is that trust could create repeat use worth earning.
A large deposit base is not a pile of idle cash
Three large banks reported the following average deposits in their consumer-focused businesses for the third quarter of 2026: JPMorgan Consumer & Community Banking, $1,078.9 billion; Bank of America Consumer Banking, $948.6 billion; and Wells Fargo Consumer Banking and Lending, $825.1 billion. Together, those reported balances are about $2.85 trillion.
These are quarterly averages for different companies’ reported segments, not a standardized household-only dataset. They include different business mixes, including some small-business relationships. They exclude other segments, such as separately reported private-bank deposits. We cannot identify from these figures how much is excess cash, low-paying, eligible for a competing account, or likely to move.
That distinction matters. Payroll deposits, pending bills, business cash and emergency reserves do different jobs. A higher rate alone does not make them equally movable.
What if banks paid an extra half a percentage point on an assumed 10% of these balances? Added annual interest would be about $539 million at JPMorgan, $474 million at Bank of America and $413 million at Wells Fargo—about $1.43 billion combined.
The same assumptions applied to 1% of the balances produce about $143 million a year; 5% produces about $713 million. These are scenarios, not forecasts. They hold balances and the extra rate constant for one full year and measure added interest expense, not changes in profit.
A bank could retain the customer by improving its offer. AI could make comparison and negotiation easier without making an established institution worthless. Keeping the account would not necessarily preserve what the bank earns from it.
Five questions every financial assistant should answer
A provider claiming to help customers choose should make five things clear:
- Scope: Which providers and products did you consider—and which could you not assess?
- Value: What would I gain after fees, eligibility rules, access needs and the costs of changing?
- Incentives: What do you earn if I follow this recommendation?
- Permission: What may you read or do, what needs my approval, and how do I stop you?
- Responsibility: Who handles a mistake, and what remedy is actually promised?
One practical test would use identical scenarios and swap provider names while keeping the relevant financial facts unchanged. The answer should follow the customer’s circumstances. If it changes, investigate why: incomplete data or different product coverage could explain the result as well as bias.
A fair test must sometimes recommend staying put. It must also ask questions when facts are missing and stop when permission runs out. We have not run that comparison inside these services, and a tool that limits itself to information should be judged against that stated role.
Different clients need different reasons to trust
A quarter-point improvement adds $125 a year on $50,000, $2,500 on $1 million, and $25,000 on $10 million, assuming an unchanged balance and rate for a year, before tax and costs. Those are cash-balance examples, not estimates of what each wealth group keeps in cash.
We use affluent for roughly $250,000 to under $1 million in investable assets, high-net-worth for $1 million to under $30 million, and ultra-high-net-worth for $30 million or more. The affluent boundary is our convention; the other thresholds follow Capgemini and exclude the main home.
An affluent household may need straightforward bill management, reliable access and affordable human help. A person with several million dollars may also need coordinated investment, tax and borrowing decisions. An ultra-wealthy family may need several specialists, legal structures, succession planning and reporting across institutions. These are service-design implications, not a claim that every client in a wealth band behaves alike.
The evidence argues against assuming that rich customers simply want one app. Capgemini’s global survey covered 6,510 high-net-worth investors. Its report says the share using a single wealth-management firm fell from 39% in 2019 to 19% in 2025. These are wealth-management relationships, not bank-switching rates or X/Revolut adoption.
An assistant might therefore earn a place by coordinating existing relationships. X or Revolut could start with a limited task and a modest balance, prove that it works, then seek broader authority. The wealthier the client’s arrangements, the more important it becomes to work with their advisers rather than ask a chatbot to replace them.
Without a branch, the fallback matters more
A familiar brand can represent years of reliable payments, established complaint routes and someone the customer expects to reach when something goes wrong. A branch is one possible form of reassurance. It is not proof of good service, and many customers can be served well without one.
J.D. Power’s 2026 study, based on 107,059 U.S. retail-bank customers, ranks trust and people ahead of saving time and money among its satisfaction dimensions. That does not measure willingness to transfer money to a fintech. It does show why a rate-only argument is incomplete.
Revolut’s UK service is built around digital access rather than a traditional walk-in branch network; X Money presents an in-app service and 24/7 support. Neither description proves the quality of help during a frozen account, a lost phone or a disputed transfer. A registered national banking branch is also different from a shop a customer can visit.
Imagine a $50,000 transfer is held for review and you cannot reach your money when needed. A higher savings rate does not answer the immediate question: who can explain the hold and help resolve it? This is a hypothetical service test, not an incident reported at X or Revolut.
A provider without a walk-in branch should show how customers recover access without their phone and reach a person with authority. Affluent clients may need dependable human help. More complex high-net-worth and ultra-wealthy clients may need a named team, access for authorized advisers and procedures for trusts, estates and succession. These are proposed requirements for winning trust, not verified X or Revolut services.
Customers also need to know which legal entity holds the money and which protection applies. U.S. deposit insurance generally covers $250,000 per depositor, per insured bank, per ownership category when that bank fails. It does not guarantee uninterrupted access or the quality of an AI recommendation. UK and European arrangements must be checked separately.
Our conclusion: let customers trust the service in stages. First prove a useful comparison. Then prove reliable access and human help. Earn permission for larger balances and more complex decisions after that.
The advantage has to be earned
Outside providers have interests too: subscription charges, referral arrangements and limited product lists can affect what they offer. A bank’s own product may genuinely be the better fit. The standard should reward a sound comparison, whoever supplies it.
The Federal Reserve’s Christopher Waller has identified open versus closed agent systems, trust and liability as important questions for commerce. The banking version adds another: will the assistant help customers choose across providers, or mainly help them use the provider that supplies it?
Schwab and JPMorgan can defend access controls and still face pressure to make approved connections useful. X and Revolut can promote choice and still face pressure to demonstrate security, clear scope and fair comparisons. Either side could earn the customer’s trust.
Watch for concrete progress: approved outside connections, reliable revocation, published comparison methods, documented cases where a provider recommends a competitor, and clear handling of mistakes. Those would tell us more than another promise of effortless AI banking.
X and Revolut could earn trust by showing when to move, when to stay, and who will help if something goes wrong. Banks could compete on the same terms. The revealing question remains: would your assistant tell me to leave when that was the better choice—and show its work?
Questions readers ask
Is Schwab banning all outside AI access?
No. Schwab describes approved data connections and plans controlled access for outside AI. Access to selected data does not automatically grant permission to trade or move money.
How much of the $2.85 trillion in deposits could move?
The public figures do not establish that. They are quarterly averages from differently defined consumer-focused bank segments, not an estimate of excess cash or balances likely to switch.
What would extra deposit interest cost these banks?
If an assumed 10% of the reported balances received an extra 0.50 percentage points for a full year, the combined added interest expense would be about $1.43 billion. This is a scenario, not a profit or switching forecast.
How can X Money and Revolut earn trust without branches?
They would need reliable access, clear fees and protections, useful human help, limited permissions and a clear remedy for mistakes. Fair comparisons should include cases where a competing provider is the better choice.
Sources and further reading
Schwab: Charley launch, September 30, 2026
Schwab: CEO interview transcripts
JPMorgan: October 13 earnings transcript, pages 8–9
Schwab: data access and customer controls
McKinsey: AI and bank customer relationships, April 30, 2026
Delegation Rights, June 2026 research preprint
Revolut: AIR launch, April 9, 2026
X Money: product information and disclosures
Cross River: X Money partnership, July 27, 2026
JPMorgan Q3 2026 supplement, page 15
Bank of America Q3 2026, quarterly segment table
Wells Fargo Q3 2026, segment deposits
Capgemini 2026 survey findings and definitions
Capgemini World Wealth Report 2026
J.D. Power 2026 U.S. Retail Banking Satisfaction Study
Revolut UK: physical branches and digital presence
FDIC: understanding deposit insurance
Federal Reserve: Payments in the Age of AI Agents, September 29, 2026
Independent analysis reviewed October 16, 2026. Deposit figures use differently defined Q3 consumer-focused segments. Interest costs are hypothetical, not forecasts. Company statements and proposed strategies are distinguished. No interviews or live recommendation tests were conducted.