Human evals for AI
A structured way to run human evals on AI products, built because usability testing alone isn't enough.
AI
10 items tagged "AI".
Case studies
Guy's and St Thomas' NHS Foundation Trust
Discovery research for 23,000 NHS staff, including the people traditional methods never reach.
Password protectedAI-moderated interviews
An AI interviewer I co-founded at Torchbox, built to reach the people traditional research can't.
Blog posts
-
AI to scale qualitative research
How we tested AI-powered interviews to reach NHS staff who didn't have time for traditional research methods.
Read → -
How we built an AI interviewer to scale user research
Under the hood of Sonar, our voice AI interviewer for user research. Co-written with the Torchbox engineering team.
Read → -
The fraud problem: why cheap research is getting more expensive
AI-moderated research promises faster, cheaper insight. But what happens when fraudsters and AI bots start talking to AI moderators?
Read → -
Researching the probabilistic product: user research in the age of AI
AI products don't stand still. User research needs to evolve from a one-off activity into an ongoing part of product development.
Read → -
The smart glasses panic
The fear of smart glasses as a category is disproportionate. Conflating specific abuses with the utility of the technology is a massive mistake.
Read → -
When AI becomes the front door to your website
AI summaries are answering for your organisation before anyone reaches your site. Usability testing wasn't built to catch what happens next.
Read → -
Why your usability testing playbook doesn't work on AI, and what to use instead
AI features don't behave like a form or a checkout flow. Testing them properly means combining human evals with automated checks.
Read →