One Knight in Product Hugo Alves - Let's Get Real About Synthetic Users (with Hugo Alves, Co-founder @ Synthetic Users)
Feb 21, 2026
Hugo Alves, co-founder of Synthetic Users and former clinical psychology master’s student, builds AI tools that generate synthetic qualitative interviews. He explains how the platform creates dynamic interview simulations, the agentic swarm of sub-agents that improves reliability, practical use in B2B research, validation methods to reduce drift, and responses to ethical critiques.
AI Snips
Chapters
Books
Transcript
Episode notes
Always Validate Synthetic Studies With Pilots
- Run many synthetic interviews rather than one to reduce the chance of a single hallucinated response skewing results.
- Hugo recommends pilots and human comparison studies so customers validate synthetic outputs against real data before trusting big decisions.
Always Define Who And What Before Running A Study
- Provide two core inputs: who you want to learn about and what you want to learn, then pick a study type (dynamic interviews, image tests, comparisons).
- Hugo notes you can either use their assistant to design recruitment and goals or skip it and supply a ready audience and question set.
Subagent Swarms Improve Quality Over One Giant Prompt
- Using a swarm of specialised sub-agents (planners, interviewers, critics) reduces drift by giving bounded tasks and incentives to each agent.
- Hugo argues task-specific agents outperform one large unfocused prompt because they avoid competing goals like persuasion vs rigor.
