SYMPOSIUM TREND LINES

BANK TO BASICS

After years of acting like tech companies, banks are keen to remind everyone they haven’t forgotten who they are. They're as determined in their pursuit of profits as ever, and AI can make them even better at it.

Banks don’t need “10,000 different AI solutions,” Citi CTO David Griffiths said during the Evident AI Symposium last Thursday in New York. Nor do they need people “scrambling around to use AI” in ways that “distract you from doing underlying process reengineering,” as JPMorganChase CFO Jeremy Barnum recently told investors.

The blinkers are on. “Some ideas try to optimize two minutes in an eight hour process and then spend millions of dollars doing that,” Bank of America CTO Hari Gopalkrishnan said. “Other ideas actually take out two steps and actually automate the heck out of three steps…doing that, you get a much better return on investment.”

For JPMorganChase, the focus is on charting the fastest route to scale: “We started a year ago what we call AI for operation, for the entire JPMorgan franchise, including the retail side, the wholesale side, across the entire thing,” said Guy Halamish, global head of digital and platform services. “We have an AI team that focused on what are the two or three big areas that we want to invest [in].” 

BMO laid out the shift to practicality most clearly: “You need to have revenue-generating use cases and get very pedantic with it,” said Kristin Milchanowski, chief AI and data officer. “[My team is] not allowed to execute any kind of use case that doesn’t drive one basis point of return…$7.6 million in net new revenue or take out $3.4 million in cost.”

“We're not doing tech for tech. We're not doing AI for AI,” David Hardoon, global head of AI enablement at Standard Chartered, told the New York audience. “It's a business objective.”

Bottom Line: The AI honeymoon is over. Investors no longer care if banks can use AI; they need to prove they can move beyond experimentation, scale it and use it to put money back in the firm’s pocket.

[ 4,200 ]

The number of “AI accelerators” at Citi, according to CTO Griffiths. These superusers share how they use the bank’s AI tools with other employees – and encourage them to use them themselves.

Why it matters: JPMC leaned on “superusers” like these to bolster AI adoption. Its LLM Suite now has more than 150,000 daily users, Teresa Heitsenrether, the bank’s chief data and analytics officer told the Symposium. JPMC employees report saving four hours per week because of its productivity features (See also “How to hook your banker on AI,” The Brief, July 10).

EARNINGS SEASON

DEEPER TALKS

The share of time during banks’ earnings calls dedicated to AI has grown fivefold compared to last year.

“That's the joke that we say: Will the last person walking out of Goldman Sachs be an engineer or a trader?”

– Marco Argenti, CIO at Goldman Sachs, at the Evident AI Symposium, Oct. 23

USE CASE CORNER

BUY, HOLD, SELL

Banks aren’t just looking to build new tools; they’re looking for better ways to use their existing ones. After Morgan Stanley scaled its AI assistant, the bank’s asset management team developed a new way to test investment ideas – which happened “just by giving them the tool,” said David Wu, head of firmwide AI product & architecture strategy.

LINCOLN LOGS

Use Case: AI @ Morgan Stanley Assistant
Vendor: OpenAI
Bank: Morgan Stanley

Why it’s interesting: The bank rolled its AI assistant out firmwide this summer, and its asset management team is using the chat feature to apply Abraham Lincoln’s “team of rivals” philosophy to investment theses.

How it works: The team tells the tool to create different personas that mirror an investment committee, developing their points of view using the firm’s internal documents and research. “They’ve got one agent that plays the bear case, one that plays the role of the bull case, one that plays risk officer and one that plays legal,” said Wu. These “agents” then debate one another about the investment’s merits and present a recommendation back to the humans in the loop.

Zoom out: Equity research is ripe for Gen AI tools and agentic forays across financial services. Manulife rolled out an AI research assistant to the firm’s wealth and asset management arm, which global chief investment officer for public markets Colin Purdie told us helps his team “understand markets better, to understand trends and data better, to go deeper in those companies” by delivering investment analysis in minutes. Blackrock, meanwhile, debuted research this summer showing how specialized agents mirroring an investment research team could “debate” one another and make a recommendation for a simulated portfolio – its own version of a “team of rivals” (see: “AI stock-pickers,” The Brief, Sept 4).

TALENT MATTERS

JPMC'S AGENTIC HIRING SPREE

JPMorganChase hired Alex Jaimes to be global head of AI agents and search, focusing on “bringing AI innovation to the products and experiences people use every day,” he wrote on LinkedIn. Jaimes was previously chief AI officer at Dataminr. The bank is also hiring an agent specialist lead on its Gen AI enablement team to develop “LLM-powered agentic systems, ensuring scalability and high performance.”

JPMC also brought on Jason Gelman as global head of AI product. He was previously in charge of the Vertex AI product at Google. Aman Thind was hired at JPMC as global head of technology strategy and enterprise architecture in London. Thind was previously CTO of State Street’s Alpha Platform.

NAB elevated Shane Conway to be its group executive, transformation, putting him in charge of the bank’s modernization and simplification agenda. He’ll report directly to CEO Andrew Irvine.

Jonathan Hollander joined M&T Bank as its head of enterprise data technology. Hollander spent the last eight months with Prudential Financial and served as corporate vice president of enterprise data platforms at TD Bank before that.

ON THE HORIZON

ACCESS GRANTED

In this new segment, we explore how a bank’s AI patent advances their AI strategy. We’ll be watching for what use cases emerge from these innovations. 

This week: Wells Fargo (fourth-largest AI patent-holder in the Evident AI Index) has a new patent that sets the stage for turning its chatbot agentic.

The patent, explained: Wells Fargo developed a new approach to security that gives the bank more control over what it lets customers do at different times. When someone interacts with the bank (say, by chatbot, as illustrated above), it uses information about their actions — like how sensitive the data they’re trying to access is or what kind of financial move they’re trying to make — to determine what kind of security check to run and what kind of access to give them.

How they can use it: It goes beyond security: These kinds of access controls go hand-in-hand with laying the groundwork for more agentic chatbots. If a bank can safely control how you interact with different systems through a chatbot, it's not hard to imagine it taking the next steps — moving money or making personalized recommendations — without ever leaving the conversation

* * * * * 

COMING NEXT MONTH: The Evident AI Patent Tracker is a comprehensive database of the AI patents filed by the 50 banks and 30 insurers we track and our analysis of the big intellectual property trends across financial services. Stay tuned.

IN THE NEWS

3 STORIES YOU NEED THIS WEEK

The AI labs’ battle for bankers is heating up: Anthropic this week launched Claude for Excel, which lets users build models inside a spreadsheet using AI, and OpenAI is contracting ex-investment bankers to create financial models it can use for training. This month, the ChatGPT-maker also bought Roi, a personal finance app that gave people tailored insights about their money. And its restructuring gives it more freedom to build enterprise products. The flurry of moves might make some banking partners start to wonder where partnership ends and competition begins (see also: “OpenAI’s Market Share Fall,” The Brief, Oct. 16). 

Agentic AI might not be the boon for business banks expect it to be, a new report from McKinsey contends. The thesis: If the public adopts tools that remove the “inertia” from financial decisions, banks stand to lose $170 billion. The solution: Banks need to build financial products agents won’t pass over.

Banks are quickly becoming campus fixtures: University partnerships are rising as banks aim to woo top talent away from the draw of Silicon Valley. This week alone, Capital One launched a new $4.5 million AI research initiative with the University of Virginia and TD Bank became a member of the MIT Media lab.

Mon 10 - 11 Nov 
FTT Fintech Festival, London

Mon 17 - Tues 18 Nov
Momentum AI Finance 2025, New York

Sun 30 Nov - Sun 7 Dec
NeurIPS, Mexico City & San Diego