Consolidating 150+ fields of client information in one click using LLM agents
WHAT Product Design, Interaction Design
WHEN 2024
Overview
When JPMorgan’s commercial bankers are preparing for client meetings, they spend hours researching their prospective client, gathering information such as the client’s financial reports, relationship with the bank, and news about their industry. They spend hours, searching across 15 different sources just to find all of this information. I led the 0 to 1 design of an internal tool that uses LLM agents to consolidate 150+ fields of client information in just one click. I led the design process from discovery to delivery – from discovery sessions with sales leads, to concept testing with bankers, to working sessions with my tech partners. I designed a quick and intuitive experience that increases efficiency for JPM bankers and allows them more time to focus on their clients.
PROJECT OVERVIEW
I conducted generative research and usability testing sessions with users.
To understand bankers’ pitch process, I designed 5 discovery sessions with sales leadership focused on what information bankers need to know about their clients as well as how they create and use client tear sheets. During these sessions, we did a cart sorting exercise to prioritize 150+ client data points, and I used our findings to design a tear sheet template. I iterated quickly, concept testing with bankers after each iteration to design an experience that is quick yet flexible to help bankers reduce time doing busy work and prioritize spending time on their clients.
I collaborated with tech partners to iterate quickly around tech limitations.
This product was the first one in JPM’s Commercial Bank to utilize LLM technology. My tech team and I worked closely together to quickly upskill on designing for and building with AI best practices. As the lead designer, I considered how this technology could optimize the tear sheet creation experience, such as opportunities to create customizable experiences for bankers. I also worked around challenges with designing for AI, such as how to be transparent with the process and establish trust in the data. I relied on the expertise of my product and tech partners as we collaborated on how to work around challenges as they arose.
Users love it: