Articos and the synthetic user research bet

Articos promises AI user research in about 30 minutes: synthetic interviews, no participant recruitment, instant insights, priced around $79 a month.

The workflow change is real. Traditional user research at an enterprise takes weeks: recruit, schedule, interview, synthesize. That latency is why so many product decisions ship unresearched. A tool that compresses the loop to half an hour does not just make research cheaper, it changes when research happens: before the roadmap discussion instead of after the feature is already committed. Having run product organizations, I would rather have imperfect signal before the decision than rigorous signal after it.

But the caveat here is the whole story: synthetic respondents are a model’s prediction of what users would say, trained on the internet’s average opinions. They are useful for pressure-testing framing, finding obvious objections, and rehearsing an interview guide. They cannot surface the surprising thing a real customer says that reframes the problem, and that surprise is the highest-value output of research. Treat synthetic research as a filter, never as evidence. The dangerous failure mode is a deck that cites “user research” without disclosing that no user was involved.

First step: run one real study and its synthetic twin in parallel, compare what each found, and write down where the synthetic version diverged. That calibration tells you which decisions this class of tool can safely own at your company.