All interviews

Brex

Staff

Own the Brex Assistant end to end — the AI agent handling real money for real employees

Brex is hiring a Staff PM to own the Brex Assistant, an AI agent that acts as an executive assistant handling expense, spend, and travel workflows conversationally over chat and SMS. The role owns product direction and strategy top to bottom, and — unusually for a PM role — goes deep into agent behavior design, prompting, eval design, and reading traces when things break, in a trust-critical financial context with no established playbook.

Step into this interview

Free · a live voice mock calibrated to this exact role

Practice this interview

What this interview tests

  • End-to-end ownership of an AI/agent product from strategy to shipped behavior
  • Eval design and feedback loops for LLM/agent reliability
  • Prompting and iterating on agent behavior, not just spec-writing
  • Operating with ambiguity as an early or sole PM on a surface
  • Setting and defending a quality bar in a trust-critical, financial context
  • Translating messy real-world workflows (expense, spend, travel) into agent capabilities

Common question themes

An AI/agent feature you shipped to real users — how you designed the evals that proved it worked

A time an agent behaved unreliably in production — how you diagnosed it from traces and fixed it

Owning a 0-to-1 or early-stage product surface with no existing playbook

How you'd define the quality bar for an agent operating with users' real financial data

Prioritizing which agent capabilities to build first given messy, high-variance real-world workflows

Working with engineering, design, and research to ship an agent-first (not screen-first) experience

View the original posting