More than 60,000 workers already have access to the LLM Suite tool, which assists them with duties including composing emails and reports.
Tens of thousands of JPMorgan Chase workers now have access to a generative AI assistant, which is the first step in a larger initiative to integrate the technology across the expansive banking behemoth.
More than 60,000 workers already have access to the LLM Suite tool, which assists them with duties including composing emails and reports. According to persons with knowledge of the plans who spoke to CNBC, the software is anticipated to become as commonplace within the bank as Zoom, the tool for video conferences.
According to the individuals, JPMorgan created LLM Suite using ChatGPT creator OpenAI’s LLM rather than creating its own AI models. LLM Suite is a gateway that enables customers to access external long language models, which are intricate algorithms that underlie generative AI technologies.
Teresa Heitsenrether, chief data and analytics officer at JPMorgan, stated in an interview that “in the end, we’d like to be able to move pretty fluidly across models depending on the use cases.” “The strategy is not to rely on a single model provider.”
The action taken by JPMorgan, the biggest bank in the United States in terms of assets, demonstrates how rapidly generative AI has taken hold in American businesses since ChatGPT’s launch in late 2022. Two OpenAI-powered products have already been made available to financial advisors by rival bank Morgan Stanley. Furthermore, significantly extending its reach, consumer tech behemoth Apple said in June that it was incorporating OpenAI models into the operating system of hundreds of millions of its devices.
JPMorgan CEO Jamie Dimon stated in April that technology, which is being heralded by some as the “Cognitive Revolution” and which would automate jobs now performed by knowledge workers, may prove to be just as significant as the invention of electricity, the printing press, and the internet.
According to Dimon, it will probably “augment virtually every job” at the bank. As of June, JPMorgan employed roughly 313,000 people.
Ban from ChatGPT
More than a year after preventing employees from using ChatGPT, the bank is now providing them with what is basically OpenAI’s ChatGPT under a wrapper approved by JPMorgan. Heitsenrether explained that this was due to JPMorgan’s desire to keep its data private from other parties.
“We don’t want our data to be used to train the model because it’s a key differentiator,” she stated. “We’ve done it so that we can use the model and maintain the security of our data at the same time.”
According to the sources, JPMorgan has implemented LLM Suite extensively throughout the organization. Groups in the investment bank, asset and wealth management, and consumer divisions of the corporation are currently utilizing it. Employees may use it to write, summarize long papers, solve problems with Excel, and come up with ideas.
However, according to Heitsenrether, who was elevated in 2023 to oversee the bank’s implementation of the trendy technology, placing it on staff members’ PCs is only the first step.
To demonstrate to others what prompt engineering is truly capable of, Heitsenrether stated, “you have to teach people how to do prompt engineering that is relevant for their domain.” “The more people delve deeper into it and discover its strengths and weaknesses, the more we’re beginning to see the ideas really taking off.”
According to her, the engineers at the bank may also utilize LLM Suite to directly integrate features from external AI models into their applications.
“Expansively larger”
After almost ten years of working on traditional AI and machine learning, JPMorgan had to change course when ChatGPT appeared.
Conventional AI, also known as narrow AI, is used for specialized tasks involving pattern recognition, such as forecasting future events based on past data. However, generative AI is more sophisticated and uses large data sets to train models with the aim of creating patterns, which is how realistic pictures or text with human voices are created.
Because LLMs are so versatile, generative AI has “exponentially more” applications than earlier technologies, according to Heitsenrether.
The bank has tested a number of these types of AI instances and has even put some of them into production.
According to her, JPMorgan is employing generative AI to produce marketing material for social media platforms, plan trips for customers of the travel company it purchased in 2022, and provide financial advisers with meeting summaries.
According to Heitsenrether, the consumer bank use AI in contact centers to assist customer care representatives in finding solutions rapidly and in determining the best locations for new branches and ATMs by absorbing satellite photos.
According to her, AI helps the company’s global payments division, which processes over $8 trillion in transactions every day, avoid hundreds of millions of dollars in fraud.
However, Heitsenrether said that due to the possibility that a chatbot would provide inaccurate information, the bank is exercising greater caution when using generative AI that directly affects a single client.
In the end, she said, the area of generative AI would produce “five or six big foundational models” that rule the market.
According to insiders discussing the bank’s AI plan, it is evaluating open source models and LLMs from US tech firms to onboard to its site next. The persons spoke anonymously.
ally or enemy?
Heitsenrether outlined three phases for JPMorgan’s generative AI development.
First, the models are merely made available to employees; second, and this is the company’s newest phase, is incorporating proprietary JPMorgan data to assist increase staff productivity.
When generative AI becomes strong enough to function as autonomous agents capable of carrying out intricate multistep tasks, it will represent a third and more significant leap that might lead to far higher productivity improvements. That would give rank-and-file workers the ability to command AI assistants, making them more like managers.
It’s conceivable that some workers will benefit from the technology while others will be replaced, altering the industry’s makeup in unpredictable ways.
According to consultancy firm Accenture, positions in banking are the most likely to be automated out of all industries, including technology, healthcare, and retail. According to Citigroup analysts, AI could increase the sector’s earnings by $170 billion in only four years.
“Like an assistant that takes away the more mundane things that we would all like to not do, where it can just give you the answer without grinding through the spreadsheets,” Heitsenrether suggested that people think about generative AI.
