JPMorgan Chase, the largest bank in the United States, has begun rolling out a generative artificial intelligence (AI) product to its employees, positioning it as a virtual research analyst.

The bank has introduced a large language model called LLM Suite to staff in its asset and wealth management division, according to an internal memo seen by the Financial Times.

The memo, signed by Mary Erdoes, head of JPMorgan's asset and wealth management business, Teresa Heitsenrether, chief data and analytics officer, and Mike Urciuoli, chief information officer of the asset and wealth management unit, described LLM Suite as a "ChatGPT-like product" for "general purpose productivity". The tool is designed to assist employees with writing, idea generation, and summarising documents.

"Think of LLM Suite as a research analyst that can offer information, solutions and advice on a topic," the memo stated. The bank has been introducing the tool to various departments since earlier this year, with approximately 50,000 employees – roughly 15 per cent of its staff – now having access to it.

JPMorgan developed the proprietary LLM in-house due to strict regulations in the financial services sector that prevent the use of consumer AI chatbots for work purposes. The bank's staff are not permitted to use external AI tools such as Anthropic's Claude, OpenAI's GPT, or Google's Gemini to ensure client data remains on secure servers.

This move represents one of Wall Street's largest implementations of an LLM, particularly one built in-house. It follows rival Morgan Stanley's partnership with OpenAI to utilise AI products for its wealth management business.

JPMorgan chief executive officer Jamie Dimon told investors in May that AI is "going to change every job". The bank's president, Daniel Pinto, has previously estimated the value from AI technology already in use at the bank to be between $1 billion and $1.5 billion.

While the rollout marks a significant step in the adoption of AI in the financial sector, it remains unclear whether LLM Suite has encountered challenges similar to other AI models, such as "hallucinating" or stating incorrect ideas as facts.

---