Lifting engagement by 47% in testing with AI-powered investing education at a leading national bank

Education was table stakes, and we were missing the mark with dense articles

CX research showed clients expected education built into the platform, not bolted on. What we had was an article farm: long, dense pieces buried away from the moments clients actually had questions.

One voice across every touchpoint, human or AI

Clients learn across experiences owned by many teams. I partnered with those teams to align on tone, terminology, and content, so AI-generated insights and editorial content feel like the same voice wherever clients find them.

Designing for the moments that drive questions

Mapping clients on a matrix, from novice to expert and self-directed to advised, surfaced the distinct needs of each user group.

Data showed most clients have advised accounts, and this same group showed the strongest demand for education. The priority was making sure AI supported their advisor relationship, not competed with it.

AI that shows up when the question does

Instead of sending clients off to find answers, LLM-powered insights surface right where questions come up, like next to a big drop on a stock chart.

Each insight is specific to the client and supports follow-up questions, going deeper without becoming a generic chatbot.

Designed to explain, never advise

In a regulated space, AI-generated advice is a risk conern. To foster innovation while working with constraints, I set a grounding principle that insights explain, never predict.

I built a repository of risk-approved topics and piloted a set that surface when their triggers are spotted, like a sharp price drop. Ask “Will it bounce back?” and users learn what drives a recovery, not a forecast.

Human in the loop

Most clients are advised, so the insights needed to be transferable to live conversations with advisors.

Insight cards for advised clients shift to show the bigger picture, and any insight can be saved with one tap to bring up in their next meeting. AI starts the conversation. The advisor finishes it.

Decoding jargon in place

Short, unobtrusive definitions sit under common investing jargon, right where clients see it. New investors get a lower barrier to entry, while experts can skim right past.

Bite-sized answers where people already look

Search is where curiosity naturally leads, so it became the home for education. In partnership with marketing and risk, long articles were condensed into bite-sized reads and videos, creating fresh content without relying on outside vendors.

From concept to funded workstream

Usability tests in Optimal Workshop showed a 47% increase in engagement with educational content. Product bought in, and the concept became a funded workstream, with a full LLM integration scoped for a future release.

Meet clients where the questions are

The most effective moments didn't feel like lessons, just answers showing up at the right time. Next steps include validating the 47% with live data and see how advised clients use insights ahead of advisor conversations.