What it does
Persona and JTBD Builder exists because of a familiar waste: a persona document written once, hung on a wall, and never opened again — because nothing in it traces back to a real customer. You supply the records — account firmographics, interview or call transcripts, support tickets, and product-usage aggregates — and the agent clusters your actual accounts into behavioural personas. Each one comes with Jobs-to-be-Done statements, verbatim evidence quotes, usage patterns and decision criteria, and every field traces back to a specific record in what you gave it. Nothing is imagined. The hard cap is the design decision worth knowing about. Past four personas a team stops using the document, so the analyzer caps output at four in code — regardless of how many clusters the data supports or how many the caller asks for. It is a refusal, not a default. Two smaller pieces of rigour stand out. The JTBD extraction distinguishes a want from a complaint: a theme whose quote negates it is reported as FRICTION rather than being flipped into a desire, and a theme with both a want and a push-back is marked CONTESTED instead of being smoothed over. And usage figures state their concentration — "630 uses across 2 of 2 accounts; concentrated, one account is 67% of it" — so a number driven by a single customer cannot pass as a shared behaviour. Every tool returns coverage warnings alongside its rows, so a persona built on thin evidence says so. Each step is callable alone: `cluster_personas` for the segments, `extract_jtbd` for the jobs, `collect_evidence_quotes` for the quotes a stakeholder deck needs, `summarize_usage_patterns`, `map_decision_criteria`. `run_full` produces the whole document. Clustering and derivation are deterministic; a language model only narrates, and quotes that look like a prompt-injection attempt are flagged rather than passed through. No connectors and no credentials — you pass the records in. Built for product managers, UX researchers and product marketers.
Example prompts
- Build personas for our Enterprise segment from this data
- What jobs are these customers actually trying to get done?
- Give me the verbatim quotes behind each persona
