AI agents as real users
Challenges of traditional research methods
Current testing methods are time-consuming, expensive and often left incomplete
Difficult to test across different user types and scenarios
Major companies (like Sonos and CrowdStrike) have faced expensive failures from bad software that impacted customers
Role & approach
Synthesised months of founder research: took all their scattered customer conversations, notes, and insights and organised them into actionable knowledge
Drafted the initial PRD: created the first coherent product requirements document that aligned everyone on scope and functionality
Used design partner insights: incorporated feedback and requirements from potential customers they were about to sign
Designed and tested a demo experience: put together simple screens for the demo experience & tested it to validate concept
Design & Implementation
Designed information architecture: structured how information and features would be organised
Created UX flows: detailed the step-by-step user interactions for core features like agent setup and test execution
Managed rapid iteration: adapted designs when built features didn't work as expected, ultimately returning to your original recommendations
Designed the Figma plugin: a separate lead generation tool that showcased the AI feedback concept
Measurable impact
5x increase in monthly active users in the first month post-launch
5-month timeline from start (December) to successful launch (May)
Rapid iteration and focused execution despite working part-time
User growth trajectory that are supporting growth conversations
Created alignment across the team on core user journey and value proposition