Major U.S. banks are moving past experimenting with AI chatbots and are now assigning autonomous "agents" real jobs inside their operations, from vetting new clients to reminding financial advisors when a customer's annuity is about to mature.
The shift, detailed in a Reuters report on Tuesday, shows firms racing to determine which functions belong to humans, which belong to software, and which are handled by a hybrid of the two.
"We are working with banks in particular on agents and human employees ... to help the banks look at all the roles end to end, and then determine which ones are hybrid roles, which ones are agentic employees, which ones are only human employees," Peter Torrente, U.S. sector leader for banking at KPMG, told Reuters.
According to a KPMG survey last month, 51% of banks were already piloting AI agents.
At Morgan Stanley, the push is aimed squarely at the advisor desk. Koren Maranca, head of artificial intelligence for wealth management at the firm, said the bank will begin testing digital assistants later this summer that interact with clients at any hour. The bank's existing agents already help advisors with routine tasks, and Maranca said the next phase involves agents that proactively "push reminders or recommendations to the financial advisors regarding their clients."
That buildout runs alongside a separate initiative Morgan Stanley disclosed in early June: opening its stock-plan administration platforms, ShareWorks and Equity Edge, to outside AI agents built by corporate clients. Mark Mitchell, chief product officer of Morgan Stanley at Work, said the firm expects corporate clients will eventually stop logging into those systems altogether, instead deploying agentic tools that pull data directly through the Model Context Protocol for connecting AI models to outside systems. The unit oversees roughly $1.2 trillion in assets and serves nearly half the S&P 500, and executives plan to extend agent access to all 3,400 stock-plan clients by next year.
At UBS, agents already generate thousands of daily alerts for advisors, flagging things such as an annuity nearing maturity that needs reinvestment. "They gather all internal information from meetings, accounts and e-mail communications," said Richard James, head of AI product at UBS, adding that once an advisor approves a transaction, the agents can execute trades and money transfers.
BNY has taken the concept furthest into the workforce itself. Digital employees there are treated as teammates assigned to specific tasks and given their own login credentials and nicknames. CEO Robin Vince described one such assistant, nicknamed Payment Pete, on a Wall Street Journal podcast earlier this year:
"The digital employee has a login, it can actually operate in the systems, and it actually has a ... human manager that's responsible for training it, making sure that it actually is doing all the right things, like a performance review, if you will, quality control, and it has tasks every day."
Behind the scenes, JPMorgan has gone further still on the investing side. A July 9 research note described eight AI-powered agents built on models from OpenAI and Anthropic that classify markets into four regimes – Goldilocks, reflation, stagflation and risk-off – and shift allocations accordingly. In backtests spanning roughly two decades, it said every agent outperformed a traditional 60/40 stock-bond portfolio on a risk-adjusted basis, with the best performer adding 0.7 percentage points of annualized return at lower volatility.
Anthropic, meanwhile, rolled out a suite of financial-services agents in early May covering pitchbooks, credit memos, KYC screening and statement audits, appearing alongside JPMorgan CEO Jamie Dimon at a New York event to promote the tools. Goldman Sachs teamed up with the AI company earlier this year to automate trading and transaction accounting, client vetting and onboarding, while JPMorgan separately sees corporate treasury as ripe for an agentic overhaul.
"Banks are increasingly using agentic AI and figuring out more ways to use it because it has a lot of potential," said Bhavi Mehta, global lead for advanced analytics in financial services at Bain & Company.
But bank investors have grown more pointed in asking where the technology spending actually shows up on the bottom line. "Investors are asking, where should we be looking for ROI on these tech spends and that's why banks are likely to focus on certain areas of AI spends where the returns are much more evident and can be scaled up," Torrente said.
The speed of adoption has drawn scrutiny from regulators and bank executives alike, given that agents are being granted access to internal systems even as oversight frameworks are still being built. Maranca said Morgan Stanley will maintain human oversight at all times and that agents will not have autonomy over portfolio decisions.
Mehta said banks broadly "are still primarily using it for internal purposes and are being extremely cautious when it touches the customer and are making sure that there is a human involved for any critical functions."
Meanwhile, Wells Fargo lures defectors from UBS and JPMorgan to expand in the East Coast, while another bank aligns itself with RayJay's financial institutions division.
New research suggests AI-exposed workers over 55 are leaving jobs more often than before ChatGPT’s rise.
Wells Fargo affiliate and women-focused wealth firm both promote leadership as they scale advisor support.
Todd Bryant of Signature Wealth Partners on vanishing pensions, SECURE Act 2.0, and what clients really want to know.
Merrill's latest hires span Colorado to Louisiana, even as industry-wide recruiting data suggests the firm is losing almost as many advisors as it gains.
Northern Trust’s Ken Lassner shows advisors how to convert volatility into after-tax portfolio gains
Dan Biagini of American Equity says the steady decline of pensions, longer lifespans and a reset in interest rates are rewriting how advisors build retirement income
