Agentic research

Agentic research moves AI from a tool used at single steps, such as transcribing this or coding those open-ends, to a system that runs the whole loop. It designs the instrument, recruits or simulates respondents, moderates, synthesises and drafts the memo. A person decides what matters and checks the work.

The failure is output nobody reviewed. The insights engineer knows where the agent breaks and puts a person there.