Building trust in Visa’s first GenAI product for clients

Visa Assist is an AI chatbot that answers business and technical questions on over 150 Visa products, so that support teams can focus on more complex cases. I led a visual redesign of Visa Assist ahead of its global launch — refining the home, conversation, and entry points to make it easy and trustworthy for enterprise users.

role: Lead Product Designer
timeline: July – Sept 2025
team: PM, Engineering, Brand

impact

Visa Assist launched globally in October 2025

Client logos: Chase, PayPal, Wells Fargo, Bank of America, MUFG, RBC, Lloyds Bank, Bradesco, and TD Bank

197
Countries
and regions
$1.2M
Support savings
in first 6 months
#1
Knowledge search
tool at Visa
74%
Response satisfaction
as usage grew 4x

problem

Visa Assist was “just another chatbot”

Chatbots at Visa have low UX consistency

Every product team was racing to build their own chatbot, each with their own look and feel. Company guidelines only covered performance, security, and accuracy. This led to duplicated efforts across engineering teams and decreasing user trust in AI responses.

Visa Assist didn’t clearly communicate its value

During pilot, employees viewed Visa Assist as a static document library, not an intelligent assistant. New users weren’t sure how it differed from other chatbots at Visa, while power users weren’t confident this would help them be self-service.

Original Visa Assist home screen, annotated with pilot feedback: “Not too confident this is self-service” and “Seems like you need to know what to do” — original home

Original Visa Assist chat screen, annotated with issues: low contrast with response, long source titles, poor readability, and unwanted rewrite options — original chat

solution

I refined Visa Assist’s most important surfaces: home, conversation, and entry points.

I worked closely with product, engineering, and brand to define requirements, co-design together, and iterate with feedback. I created a prototyping workflow in Claude Code with product context, design system, and user research findings built in.

Redesigned Visa Assist home screen with a personal greeting and six categorized conversation starters — redesigned home

Redesigned Visa Assist chat screen with a structured, numbered answer followed by source cards — redesigned chat interface

Marketing graphic on a laptop lock screen announcing that Visa Assist is now available to all clients — marketing graphics

Reflection

What I learned designing AI for trust, at scale

Think systems, not screens

Clients have many touch points with Visa. My job wasn’t just making the product great, but also meeting clients wherever they are in their support journey.

Adoption across design

If I had more time, I wish I brought the rest of our design organization along for the ride. I’d host more share outs with design teams building chatbots.

The Visa Assist team gathered for a group selfie in the office