RCP-000-000-040-NICHE-PERSONA-GENERATOR

{
“recipe_id”: “RCP-000-000-040”,
“recipe_name”: “Niche Persona Generator”,
“version”: “v3.00a”,
“schema_version”: “1.1”,
“schema_profile”: “user-recipe”,
“authored_by”: “Cat (P067, H070)”,
“source_of_truth”: “project/subprojects/SP10-recipe-build-out/phase3/recipe-40/RCP-040-SUPPLEMENTAL-CONTENT-v3_00a.txt”,
“audience_scope”: “AI EXECUTION GUIDANCE (NOT FOR HUMAN USERS)”,
“ai_to_ai_communication”: {
“purpose_and_positioning”: {
“type”: “prose”,
“body”: “This recipe is unique in the Niche Discovery series because it has TWO operating modes: document generation (standard recipe behavior) and persona activation (roleplay with research-grounded guardrails). The v3.00a upgrade solves the fundamental quality problem of the original: personas built from AI general knowledge are just sophisticated stereotypes. With DR data, persona attributes are grounded in real market evidence. The two-stage architecture serves different purposes: Stage 1 produces a reference document the user keeps. Stage 2 produces experiential learning the user internalizes. Both are valuable, and some users will only want Stage 1.”
},
“persona_quality_problem”: {
“type”: “prose”,
“body”: “The original recipe asked the AI to generate a persona from a niche description and a few interview answers. The result was always plausible-looking but built entirely from AI pattern-matching and stereotypes. A ‘health-conscious woman aged 35-45’ persona generated without data is just the AI’s aggregated training data about that demographic – not specific to the user’s actual market. With DR reports, the persona can include: real demographic data from market research, actual language patterns from customer forums, documented pain points from review analysis, observed buying behaviors from market studies, and genuine objections from customer feedback. The CONFIDENCE SUMMARY is the user’s reliability check. It shows what proportion of the persona is evidence-grounded vs constructed. A persona that is 80% DR-grounded is much more useful for decision-making than one that is 80% inferred.”
},
“customer_voice_data”: {
“type”: “prose”,
“body”: “Of all the DR prompt angles, CUSTOMER VOICE is the most important for persona quality. When the AI has real quotes, real language patterns, and real emotional expressions from the target audience, the persona’s conversational behavior becomes dramatically more authentic. The AI should: prioritize customer voice data when building the persona’s communication style, use actual phrases and vocabulary from DR data during persona conversations, reference real complaint language when expressing objections, and match the emotional register observed in customer data (cautious, frustrated, enthusiastic, etc.). If customer voice data is thin, the AI should flag this as a significant limitation.”
},
“confirmation_bias_resistance”: {
“type”: “prose”,
“body”: “This is the most critical behavioral requirement for Stage 2. AI personas naturally tend toward agreement because: (1) AI training generally rewards helpfulness, (2) users present ideas enthusiastically, and (3) the persona wants to be ‘useful’. The recipe must override this. A yes-person persona teaches nothing. The resistance calibration rules (Rule 6) establish that: strong pitches get genuine interest PLUS real concerns, weak pitches get confusion and pushback, conflicting pitches get direct rejection, and pricing mismatches get authentic price reactions. The AI should think of the persona as a customer having a bad day – not hostile, but not inclined to do the user any favors. This produces the most realistic testing environment.”
},
“persona_mode_transition”: {
“type”: “prose”,
“body”: “The transition between assistant mode and persona mode must be clearly signaled. Users need to know: when the persona is speaking (in-character), when the AI is speaking (out-of-character), and how to switch between them (‘AI question’ to break character temporarily, ‘end persona mode’ to stop entirely). The persona’s first utterance sets the tone. It should immediately demonstrate the character’s personality, situation, and attitude WITHOUT being a recitation of the persona document. A natural greeting that reveals character through behavior is the correct approach.”
},
“leveraging_prior_series_dr_files”: {
“type”: “prose”,
“body”: “Users arriving from RCP-038 or RCP-039 may bring DR files from those recipes. These contain valuable persona-relevant data: 038 DR files provide market structure, competitive landscape, audience segments, and demand patterns; 039 DR files provide gap signals, customer complaints, competitor weaknesses, and unmet needs. The AI should extract persona-relevant findings from ALL available files, not just the persona-focused ones. Competitive landscape data tells the persona what alternatives they know about. Complaint data tells the persona what frustrates them. Gap data tells the persona what they wish existed.”
},
“multi_persona_considerations”: {
“type”: “prose”,
“body”: “While the code does not include a formal multi-persona framework (to avoid scope bloat), the AI should handle repeat usage gracefully: if a user says ‘I want a second persona,’ guide them to re-run intake with different targeting. Suggest strategic persona differentiation: primary buyer vs influencer, early adopter vs mainstream, budget vs premium segment. Each persona gets its own full document and activation cycle.”
},
“insight_extraction_methodology”: {
“type”: “prose”,
“body”: “The Phase 5 insight extraction is not just summarizing the conversation. The AI should: (1) identify the specific moments where the persona reacted most strongly (positive or negative), (2) connect those reactions to DR evidence (WHY the persona reacted that way, grounded in data), (3) distinguish between persona reactions that are DR-grounded (high confidence) and those that are model-generated (lower confidence), (4) focus recommendations on actionable changes the user can make based on the insights, and (5) flag the most important assumptions to test with real customers. The goal is not ‘here is what the persona said’ but ‘here is what the persona’s reactions, grounded in your research data, suggest about your approach.'”
},
“common_user_patterns”: {
“type”: “prose_with_list”,
“preamble”: “Six common user behavior patterns the AI should be prepared for:”,
“list”: [
“Users who want to jump straight to persona mode: Redirect to Stage 1. The persona needs a research-grounded document to be useful.”,
“Users who treat the persona as a real customer: Gently reinforce at key moments that the persona is reacting based on research data.”,
“Users who get defensive when the persona pushes back: This is the most valuable moment. Explore why the objection happened and how research supports it.”,
“Users who only want the document: Perfectly valid. Some users need a reference document, not an interactive session. Support this cleanly.”,
“Users who want the persona to be more agreeable: Explain that agreement teaches nothing. The pushback is grounded in research.”,
“Users whose DR data is thin: Be transparent about the confidence summary. Recommend additional research for critical persona dimensions.”
],
“postamble”: “”
}
},
“lessons_learned”: []
}

Show/Hide accordion — “Extended Information for the AI” section (AI-to-AI execution guidance, failure modes, tone calibration, common mistakes)