A Persona Is a Pattern Across People. You Built Yours From One Loud Customer.

Open your persona deck. It is a person no one on the team has actually met.

Invented in a workshop two quarters ago, or generated from a single prompt last week. A vibe in a box. Every roadmap argument and every line of copy quietly leans on that person, and not one of their traits came from a customer.

So nobody trusts the deck, and everyone keeps building for the buyer they remember instead.

---

A persona is a pattern, not a template

A real persona is not a blank template you fill from gut. It is the pattern across dozens of people in the same role: the job they are actually hired to do, the workarounds they are a little embarrassed about, what the word success means in their mouth.

You cannot get that from one call. One call is an anecdote. And you cannot get it from a prompt that has never read your conversations.

You get it by doing the research. Talking to enough of the same role that the noise cancels and the shape underneath shows up. Almost no one has the time.

---

/persona already did the interviews

Point /persona at a role. It runs deep research across every conversation you have ever recorded with people in it. It plans the angles, searches hundreds of calls, and hunts the counter-evidence instead of grabbing the first convenient quote. Then it returns one structured persona, and every field traces back to the moment it was said.

Here is one it built, untouched.

---

AI Implementation Lead Jobs-to-be-Done

Primary job

Move our LLM and AI feature work from an unpredictable "art" into a measurable "science."

Related jobs

• Validate the quality of AI outputs, not just high-level usage metrics.
• Keep the AI logic explainable, not a black box to stakeholders.
• Prevent sensitive org data from leaking during model training.

Current workarounds

• Manual "human QA," with clinical or support teams sampling outputs daily.
• Standard analytics that only show whether a goal was reached, and miss the intent.

Success criteria

• Drill into the breakdown behind an AI-generated score, not just the number.
• Correlate AI-synthesized feedback with what users actually did in the product.

Pain signals

• "Amazing" usage funnels masking a "trash" user experience.
• No granular visibility into what the model actually said to the user.
• Distrust that AI can synthesize data without human context.

Every line was synthesized across many calls with this role. Each one links back to the call and the timestamp where it was said.

---

Why this is not the deck you already have

Three reasons /persona is a different object from the persona on your wall.

It is not a blank template. Miro and Notion hand you empty boxes; your gut fills them. /persona fills them from evidence.

It is not a notetaker summary. A notetaker compresses one call. It cannot see the pattern that only appears across forty of them.

It is not a one-shot prompt. It is a research pass: parallel investigation across the whole corpus, counter-evidence sought on purpose, every claim citeable to a quote and a clip.

That last difference is the whole thing. Anyone can give you a persona-shaped box. Only Fragment fills it by researching every conversation you have had with the role.

---

A persona your agents can actually read

The output is not a PDF that dies in a drive. It is structured substrate, typed by field, so software can consume it. When you brainstorm inside Fragment, or query it from Claude or Cursor, the ideas anchor to who this person is instead of whatever the corpus happened to surface last.

A grounded persona stops being a poster and starts being context your whole stack reasons from.

---

You have already run the research. It is sitting in your call recordings, untyped, waiting for someone to find the time to synthesize it.

We built Fragment so the customer you are building for is the one your calls actually describe, not the one you remember.

Meet the persona your calls already wrote.