Top Midday Stories: Apple to Invest $500 Billion in US; Apollo Acquires Bridge Investment for $1.5 Billion

By Alexander Saeedy

America's banks have been using artificial intelligence to spot fraud for years. JPMorgan Chase, the country's biggest bank, is now making a bigger bet on AI, working to put it at the center of how its 300,000 employees work.

In the past year, JPMorgan rolled out a tool it calls LLM Suite for most of its employees that allows them to use generative artificial intelligence from OpenAI and others. It's also adding generative AI in call centers for agents that deal with Chase Bank customers.

Teresa Heitsenrether, JPMorgan's chief data and analytics officer, was tapped to oversee the firm's AI strategy in 2023. She isn't a software engineer or a technologist. But after more than 20 years running key businesses at JPMorgan, she was able to identify where the firm can best use technology to boost productivity — and had earned the trust of the company's leaders.

"When the generative AI wave started to really take off, we wanted to make sure that there was somebody really focused on this as a full-time job at [CEO] Jamie Dimon's table because of the strategic importance that we see for the future of the firm," she says.

Heitsenrether talked about where she sees AI going at JPMorgan in the next few years.

How often are employees using the LLM Suite?

We now have 200,000 people in the firm who have the tool on their desk. Half of them use it pretty actively every day, and we estimate people use it on average for one-to-two hours a week.

How do you think generative AI will transform work at JPMorgan?

Think about anyplace in the bank where people are preparing to go and talk to their clients. Today, you have armies of people running around, pulling briefing memos together and making sure that everybody's prepped. This is a great way of being able to pull those things together more quickly. We see it in legal, in any place where you've got lots of documents, lots of information to sift through. It's a real time saver there. And in the call centers, too.

What are you seeing in the call centers?

They've been using AI for a while to help answer customers' questions. But when you couple what we already used with the new large language models, they're great. Think about a consumer calling. It might be about your car loan, your mortgage, your deposit account, your credit card. The person on the other end of the phone has to be able to traverse all those different products. The models let the call-center agent find information faster and answer the questions faster and more accurately.

Is there any thought about developing your own large language models?

Probably not. You look at the scale of these things and what they're trained on, and the cost. There are a number of them that you can work with, and they're all pretty darn impressive. Building our own model isn't how we differentiate ourselves. The secret sauce for us is how you use them with our information.

Are you seeing tangible productivity gains across the bank from AI?

It's very early innings. First we want to put the tool in people's hands, and let them be able to ask questions and get answers. That already starts to spawn ideas, innovation, some productivity.

The second phase is where you take the models and start to plug them into JPMorgan: our policies, our procedures, what we know about clients. Now you can have the model working on knowledge that's specific to us.

The third horizon, and we're not there yet, is for the models to be able to do more reasoning. What happens is they get a chance to think, OK, based on the complexity of the problem that you're asking me, let me think in the same way a human being would approach it. It lets the model find the resources it needs. Maybe it's going to go to the internet or some system outside JPMorgan's databases. You can effectively take the workflow of somebody who's an investment-banking analyst or client-service person and teach the models the steps they would take to get their jobs done.

We will always have a human in the loop to check the models' work. I don't think we would, certainly not at this juncture, let these things be autonomous.

How do you think about the impact of AI on the structure of the workforce?

It's too early to tell where the impacts are going to be. I think there will be impacts. I don't think that this is the death knell or elimination of loads and loads of people. Unlike a lot of prior waves of technology that have been all about automating tasks — and they tend to be more clerical repetitive tasks — this is applicability across the board. There aren't a lot of job functions that can't be augmented with this tool. So we see it as a combination of humans and [AI] models, being able to do more.

What innovations in your technology do you think will have the most impact?

The models are all trained on all the data that's publicly available, right? What the models don't have access to is the data that's specific to the enterprise. That then becomes the differentiator in terms of how much value JPMorgan gets versus somebody else in financial services. A lot of the focus we have is on how do you safely and effectively leverage the knowledge and the insights that we have in the firm.

Have you already started to connect the models to JPMorgan's proprietary data?

We are starting. It's a work in progress. We're obviously doing that very carefully and thoughtfully. Data privacy is our first and foremost priority.

How do you make sure that your AI isn't running amok?

We have controls around everything we've done with AI that have been long established. Every time we develop a use case, it goes through a rigorous process. Some models we've designed are very constrained. You can see everything it's doing. So that might be something you would use in a credit decision.

Then there are the large language models. They are more of a black box. You're never going to use that for that purpose. You might use it to create a travel itinerary for somebody from the Chase Travel side, because there's not a high-risk scenario. So it's a question of having the right tool in the first place and then having the right controls.

How are JPMorgan clients able to access the bank's AI tools for themselves?

AI does give you the ability to scale and provide more personalization. Most of us, I think, would prefer if you can just go online to a digital channel and get the answer fast. The AI can help to discern your question and get you an answer in a digital channel more quickly.

Is there any specific area where you're already seeing good returns on your AI investment?

I would turn it around. I can't think of a business where that's not true.

Interview has been condensed and edited.

Write to Alexander Saeedy at alexander.saeedy@wsj.com