Skip to content

Personas — Synthetic User Research

active

AI-powered synthetic user profiles that simulate real user segments. Interview them, run Kano surveys through them, simulate product scenarios — calibrated by Harlan's conversations, Wren's research, and Margot's market data.

observatorypersonassynthetic-researchaiinterviewsscenarios
See also: overviewcrew › margotcrew › wrencrew › harlan

Personas — Synthetic User Research

See also: overview.md | crew/margot.md | crew/wren.md | crew/harlan.md

“I’m the only one who talks to real people. That either makes me the most important person on this crew, or the most exhausted. Most days, both.” — Harlan


OwnerMargot
MCP@sector137/mcp-observatory
Slash/observatory
CLIsector137 observatory personas
SDKsdk.observatory().listPersonas()
Statusactive

TL;DR

  • Personas are AI-powered synthetic user profiles that simulate real user segments
  • Three capabilities: AI Ask (in-character interviews), Synthetic Kano Survey (feature evaluation), Scenario Simulation (product reaction)
  • Not replacements for real research — augmentations. Force multipliers for early-stage hypothesis testing
  • Calibrated by three signal sources: Harlan’s customer conversations, Wren’s user research, Margot’s market data
  • Personas get sharper over time as the crew feeds them real-world signal
  • Linked to roadmap issues via personaIssues junction — “which personas care about this issue?”

What Personas Are

Personas are synthetic humans the crew builds to simulate real user segments. Each persona has a role, goals, pain points, technical level, communication style, and enough context to respond in-character to questions about your product.

They are not stock photos with job titles. They are AI-powered simulations calibrated against real customer data. When Harlan talks to a real Series A CTO and hears what keeps them up at night, that signal gets folded into the synthetic CTO persona. When Wren observes how a specific user segment navigates a prototype, that behavioral pattern refines the persona’s experiential profile. When Margot reads a market shift that changes what a segment values, the persona’s strategic context updates.

The personas are living documents with opinions. Ask them a question and they answer in-character. Run a survey through them and they respond based on their simulated preferences. Describe a scenario and they react with structured feedback — sentiment, insights, concerns, alternatives they would consider.

Status: LIVE. Full implementation: personas table, /api/v1/personas CRUD routes, AI conversation engine (persona-ai.ts via Vercel AI SDK + Anthropic Claude Sonnet), synthetic Kano survey, scenario simulation, persona-issue linking via personaIssues junction table. 9 MCP tools. UI: personas.tsx (list), persona-detail.tsx (profile + AI ask/survey/scenario tabs).


Three Capabilities

AI Ask — In-Character Interviews

Chat with a persona about your product. They respond in-character based on their full profile — role, goals, frustrations, technical level, communication style.

This is not a chatbot answering as “a generic user.” The persona carries the accumulated context of their simulated segment. Ask a synthetic DevOps lead about your deployment feature and they respond from the perspective of someone who has dealt with broken CI pipelines, 3 AM incidents, and the specific frustration of tools that solve the wrong problem.

Use this for rapid hypothesis testing: “Would this persona care about feature X?” “How would they describe this problem in their own words?” “What would make them switch from their current solution?”

Synthetic Kano Survey — Feature Evaluation

Run a Kano study questionnaire through a persona’s simulated preferences. The persona evaluates each feature’s functional and dysfunctional dimensions as if they were a real survey respondent from their segment.

This feeds directly into Intel Missions (see intel.md). Synthetic Kano responses supplement real human responses — useful when you need preliminary signal before investing in survey distribution, or when you want to stress-test assumptions about how different segments would classify the same features.

Synthetic data is always marked as synthetic. It augments, never replaces.

Scenario Simulation — Product Reaction

Describe a product scenario and get the persona’s structured reaction. The output includes:

  • Sentiment — positive, negative, neutral, or mixed
  • Key insights — what resonated, what fell flat
  • Pain points — concerns or objections the persona would raise
  • Likelihood to adopt — a simulated signal, not a guarantee
  • Alternatives considered — what the persona would compare your scenario against

Scenario simulation is Wren and Harlan’s primary tool for testing experience hypotheses. Before committing to a design direction, run the scenario through three personas from different segments. If all three light up, the signal is strong. If one lights up and two shrug, you have a niche feature. If none care, the scenario needs rethinking.


Who Calibrates Them

Personas are only as good as the signal that shapes them. Three crew members contribute distinct calibration streams:

Harlan — Ground Truth

Harlan is the only crew member who physically crosses to the human world. His customer conversations are the raw material. When he talks to a real person and hears what they actually care about — not what they say in a survey, but what they reveal between the lines — that signal calibrates the personas.

Harlan’s conversations set the persona’s emotional truth: their real frustrations, their actual decision-making process, the things they would never write in a form but will say over coffee.

Wren — Experiential Lens

Wren watches how humans interact with products. Her research refines the persona’s experiential profile — how they navigate, where they get stuck, what makes them feel competent versus confused. Wren does not translate user behavior into metrics. She translates it into understanding.

When Wren says “this persona would abandon the onboarding at step 3,” she is not guessing. She is applying observed behavioral patterns to the persona’s profile.

Margot — Strategic Context

Margot shapes the persona’s market context. What does this segment value today? What will they value in six months? Are they price-sensitive or value-driven? What competitive alternatives are they evaluating?

Margot’s Intel Mode feeds the persona’s strategic layer — the context that determines not just whether they would use a feature, but whether they would pay for it, and how much.


Persona-Issue Linking

Personas can be linked to roadmap issues via the personaIssues junction table. This answers a practical question: “Which personas care about this issue?”

When the crew is prioritizing the backlog, persona links add a human dimension to the decision. An issue linked to three high-value personas carries more weight than one linked to none. An issue that no persona cares about is a candidate for the Indifferent quadrant.

Persona links are also useful in reverse: viewing a persona’s profile shows which issues affect them, giving the crew a segment-level view of the roadmap.


MCP Tools

ToolOwnerPurpose
list_personasMargot / Wren / HarlanList all personas — filter by status or search by name
get_personaMargot / Wren / HarlanFetch a single persona with full profile
create_personaMargot / Wren / HarlanCreate a new synthetic user persona
update_personaMargot / Wren / HarlanModify persona fields — demographics, goals, frustrations, technical level
delete_personaMargot / Wren / HarlanPermanently remove a persona
ask_personaWren / HarlanAI interview — persona responds in-character to a question
run_persona_surveyMargotSynthetic Kano survey — persona evaluates features as a simulated respondent
run_persona_scenarioWren / HarlanScenario simulation — structured feedback including sentiment, insights, pain points
list_persona_conversationsMargot / Wren / HarlanConversation history for a persona — all past Q&A sessions

Events

EventTriggered ByActorDownstream
persona.createdUser creates a personaCaptain / crewFeed update, Library research archive
persona.updatedPersona fields modifiedCaptain / crewFeed update
persona.deletedPersona permanently removedCaptainFeed update, Library archive pruned
persona.question_askedAI ask session runsWren / HarlanLibrary: conversation history updated
persona.survey_runSynthetic Kano survey completesMargotLibrary: survey results stored, Kano study can pull synthetic data
persona.scenario_runScenario simulation completesWren / HarlanLibrary: scenario feedback stored

Implementation

PackageDescription
apps/appCore system — persona AI engine (lib/persona-ai.ts), API routes (routes/api/v1/personas.ts), client pages (personas.tsx, persona-detail.tsx)
packages/sdkTypeScript SDK — persona resources (CRUD, ask, survey, scenario)
packages/mcpMCP server — 9 persona tools