Next

Bonnie Chen

Designing an AI agent that flags errors for TurboTax’s professional tax preparers

Long-term goal: bring quality review down to ~25 minutes per return. MVP target: save 3.6+ minutes per return, about $6M in projected savings.

QR Agent checking a return for issues
Role
Product designer
Timeline
October 2025 – February 2026
Status
Piloted on ~1,500 Schedule C returns · 82.4% precision

At a glance

QR Agent checks tax returns for errors, so experts only spend time reviewing and resolving rather than looking for errors.

Problem
Only 23% of returns get a quality review, each takes about an hour, and over 80% of reviewed returns have no errors.
My role
I led the design of this agent experience with a five-person team, driving prototypes from first concepts to a live pilot.
Team
Product designer (me) · Product manager · Two tech leads · User researcher
Scope
Build an end-to-end agentic UX within the existing tax expert platform, IEP.

Please reach out for access.