Product selection agent for technical B2B sales

Product selection agent for technical B2B sales

An agent that helps sales reps find the right product part out of thousands in seconds.

An agent that helps sales reps find the right product part out of thousands in seconds..

An agent that helps sales reps find the right product part out of thousands in seconds..

Role

Product designer

Owned

Agent UX, behavior, UI

Company

Rapidflare.ai

Timeline

Apr 2025 – Jun 2025

Challenge

Sales reps in semiconductors have to match customers to the right part across thousands of options and dense specs. This process spans hours and even days, since most sales reps don’t have the technical depth to do it alone. Rapidflare had a working Product Selection Agent that could already narrow to a recommendation, but it did not handle ambiguous input well.

My role: The experience layer. I owned agent behavior, response architecture, communication style, and the UI. I worked in weekly cycles with product and engineering, and tested directly with senior sales engineers at early customers.

The goal: Not just accurate recommendations, but an agent that’s honest about what it assumed, and easy to correct when it’s wrong.

The solution

User initiates and agent extracts intent

User initiates and agent extracts intent

Agent narrows down thousands of parts

Agent narrows down thousands of parts

Agent recommends and flags assumptions

Agent recommends and flags assumptions

Outcome

  • Exceeded the accuracy bar for major early customers

  • 80% of pilot users used filters mid-chat

  • Behavior model became the standard across all agents

  • Launched publicly at GSX 2025

How this got built

Most of the behavior model came out of testing. Many of the paths apart from the happy path, like when a rep has a number but is just trying to assess feasibility, were surfaced in working sessions with product and engineering. Testing with sales engineers exposed new use cases and failure modes as well.