Skip to content

banking · production case

JPMorgan Chase

LLM Suite and legal agentic workflow

The largest public employee-facing agent platform in banking, now being extended into multi-step legal and research work.

What it does

JPMorgan rolled self-service access to its LLM Suite to about 200,000 employees in under a year; Accenture reports that half of those users open it three or more times a day. A separate Legal Agentic Workflow (LAW), built with AWS, OpenAI and Snowflake, was reported to determine a contract termination date accurately more than 95% of the time, versus under 3% for GPT-3.5-Turbo on the same task.

Architecture

A governed model registry with split paths: sponsor review for low-risk uses, a full model-risk committee for higher-risk ones. Specialist agents sit behind the employee channel rather than a public chatbot.

Evidence

Employee scale from Accenture’s banking blog (January 2026). LAW accuracy from Evident’s “Banks go agentic” briefing. CEO Jamie Dimon has said there will be “no job, no process, no function that won’t be affected by AI.”

Caveat

LLM Suite is an internal copilot, not a customer-facing autonomous agent. LAW is a documented workflow with a published accuracy claim, not a fully unsupervised legal actor.

Sources

Go deeper

Four briefings, grounded in this case and the rest of the observatory. Not a chatbot reciting the page.

← All production cases