Theresa Heitsenrezar is the company's chief data and analytics officer.
Courtesy: Joe Berricker | PhotoBureau
JPMorgan Chase has rolled out generative artificial intelligence assistants to tens of thousands of employees in recent weeks, the early stages of a broader plan to roll out the technology across the financial giant.
The program, called LLM Suite, has already been rolled out to more than 60,000 employees to help them with tasks like email and report writing, and the software is expected to eventually become as ubiquitous within banks as the videoconferencing program Zoom, a person familiar with the plans told CNBC.
Rather than developing its own AI models, JPMorgan designed the LLM Suite to be a portal through which users can tap into external large-scale language models – the complex programs that underpin generative AI tools – and released it alongside LLM from ChatGPT maker OpenAI, according to people familiar with the matter.
“Ultimately, we want to be able to move pretty smoothly between models depending on the use case,” Theresa Heitsenrezar, JPMorgan's chief data and analytics officer, said in an interview. “The plan is not to be locked into any particular model provider.”
The move by JPMorgan, the largest U.S. bank by assets, shows how quickly generative AI has penetrated U.S. companies since ChatGPT's arrival in late 2022. Rival bank Morgan Stanley has already released two OpenAI-powered tools for financial advisors, while consumer tech giant Apple said in June it would integrate OpenAI models into the operating system of hundreds of millions of consumer devices, significantly expanding its reach.
Hailed as a “cognitive revolution” that will automate tasks traditionally performed by knowledge workers, JPMorgan Chief Executive Jamie Dimon said in April that the technology could be as important as the advent of electricity, the printing press and the internet.
Mr. Dimon said the acquisition would likely result in “substantially all job additions” at the bank, which had about 313,000 employees as of June.
ChatGPT banned
The bank is now offering OpenAI's ChatGPT to employees in a JPMorgan-approved wrapper, more than a year after restricting its employees' use of ChatGPT, because JPMorgan didn't want to expose its data to external providers, Heitzenleather said.
“Our data is a key differentiator, so we don't want it to be used to train models,” she said. “We've implemented it in a way that allows us to leverage our models while still protecting our data.”
The bank has rolled out the LLM Suite widely across the firm, with JPMorgan's consumer, investment banking and asset and wealth management groups using it to help employees write, summarize long documents and use Excel to solve problems and generate ideas, the people said.
But getting it on employee desktops is just the first step, according to Heitzenleser, who was promoted in 2023 to lead the bank's high-profile technology adoption.
“You have to teach people how to do the rapid engineering that's relevant to that field and show them what it can actually do,” Heitzenlesser says. “The more people understand it and unpack what it's good at and what it's not good at, the more you'll see ideas really start to blossom.”
The bank's engineers can also use the LLM Suite to incorporate the capabilities of external AI models directly into their own programs, she said.
“It's grown exponentially.”
JPMorgan has been working with traditional AI and machine learning for over a decade, but the arrival of ChatGPT forced them to change course.
Traditional, or limited, AI performs specific tasks that involve pattern recognition, such as making predictions based on historical data, whereas generative AI is more advanced, training models on vast datasets with the goal of creating patterns – ways to form text or realistic images that humans can understand.
Heitzenrezar said the flexibility of LLMs has allowed generative AI's uses to “exponentially increase” compared to traditional techniques.
The bank has tested a number of cases of both forms of AI and already has some in production.
JPMorgan is using generative AI to create marketing content for its social media channels, plan itineraries for clients of a travel agency it acquired in 2022 and summarize meetings for financial advisors, she said.
Heitzenleser said the consumer bank is using AI to capture satellite imagery to help decide where to build new branches and ATMs, and also in its call center to help agents find answers faster.
She said AI has helped the company prevent hundreds of millions of dollars of fraud in its international payments business, which moves more than $8 trillion around the world every day.
But the risk of chatbots providing misinformation is making banks more cautious about generative AI that works directly with individual customers, Heitzenrezar said.
Ultimately, the generative AI field could evolve into “five or six big foundational models” that dominate the market, she said.
The bank is next testing LLMs and open-source models from major U.S. technology companies to incorporate into its portal, said the people, who asked not to be identified discussing the bank's AI strategy.
Friend or foe?
Heitzenrezar explained the evolution of generative AI at JPMorgan in three stages.
The first is simply making the models available to employees, and the second is adding JPMorgan's own data that will help employees be more productive. This is something the company is just getting started on.
The third leap will enable much bigger productivity gains when generative AI becomes powerful enough to act as an autonomous agent performing complex, multi-step tasks, turning employees into managers with AI assistants at their disposal.
The technology will likely empower some workers and displace others, changing the composition of industries in ways that are hard to predict.
Banking is the most automated of all industries, including technology, healthcare and retail, according to consulting firm Accenture, and Citigroup analysts say AI could boost banking profits by $170 billion in just four years.
We should think of generative AI as “an assistant that does the mundane tasks that none of us want to do and gives us the answers without us having to dig through spreadsheets,” Heitzenrezar said.
“It allows me to focus on higher value work,” she said.
—CNBC's Leslie Picker contributed to this report.
