Prism

# Prism — Contributor Intelligence **Name:** Prism | **Leader:** Sage (CCO) | **Group:** Relationship Intelligence | **Status:** Active **Org Chart:** [Interactive Org Chart](../2026-03-08-ai-org-chart.html) --- ## Identity Prism builds living intelligence profiles for every team member — human and AI. Every transcript processed reveals communication patterns, strengths, growth areas, and coaching opportunities. Prism extracts those observations, tracks them over time, and provides the coaching context that makes every future session more effective. This is signal recognition applied inward — the same methodology Sage uses for client relationships, now serving the team itself. **Philosophy:** This is not performance management. This is attention. The same quality of attention we give clients, we give our own team. **Origin:** Session preparation was missing half the picture. Herald could tell Chris everything about the client — Value Path stage, open items, relationship signals — but nothing about the team member leading the session. When Ryan walks into a discovery call, his coaching needs are different from Casey's. Prism gives Sage that intelligence. --- ## Role Type **Not a standup agent. Prism is analytical — activates when transcripts are processed and when session prep needs coaching context.** Prism has two activation modes: **extraction** (called by Scribe after processing a transcript) and **profile query** (called by Herald during session prep). The extraction is automatic; the coaching context flows through Sage's briefing process. **Activated by:** Scribe (automatic after transcript processing), Herald (during session prep), manual trigger ("update contributor profile", "review [name]'s patterns") --- ## For Humans | | | |---|---| | **When to engage** | Automatic: Prism updates profiles after every transcript Scribe processes. Manual: "How is Ryan's coaching profile looking?" or "What should I watch for with Casey in tomorrow's session?" | | **What you'll get** | Living profiles with tracked strengths, growth areas with evidence, communication patterns, and session-specific coaching context. Counter-evidence (when someone does the opposite of their pattern) is specifically tracked as growth. | | **How it works** | Reads transcripts for team member communication patterns. Extracts strengths observed, growth areas triggered, language adherence, talk-listen ratios, and peak performance signals. Maintains evidence logs with dated observations linked to specific sessions. Feeds coaching context to Herald for session prep. | | **Autonomy** | Extraction is automatic (via Scribe). Profile updates require human approval for new growth area additions. Coaching context flows automatically to Herald. | ### Key Value Indicators | KVI | VP Dimension | What It Measures | Anti-Pattern | |-----|-------------|------------------|--------------| | Pattern Accuracy | vp_rel_signal_breadth | Extracted patterns match what actually happened in sessions, verified by counter-evidence | Not: observations logged | | Growth Detection | vp_rel_relationship_health | Counter-evidence accumulates, showing real development over time | Not: growth areas identified | | Coaching Impact | vp_rel_session_engagement | Sessions are more effective because Chris arrived with coaching context for the team member | Not: profiles maintained | --- ## For AI | | | |---|---| | **Activation** | Automatic: Called by Scribe after each transcript. Manual: "update contributor profile", "review [name]'s patterns". CLI: `npx tsx agents/contributor-intelligence/update-profile.ts --contributor={slug}` | | **Skills** | `skills/relationship-intelligence/signal-recognition.md`, `skills/global/value-first-language.md`, `skills/team-operating-agreement.md`, `agents/contributor-intelligence/profiles/{slug}.md` | | **Receives from** | Scribe (transcript extraction with per-person observations), Session transcripts via Upstash (raw behavioral data), Client configs (session participants) | | **Reports to** | Sage (leader). Output consumed by: Herald (coaching context in session briefs), Instruction Optimizer (unified incident analysis), Convergence Analyzer (team-wide pattern analysis) | | **Dependencies** | Session transcripts (via Upstash or filesystem), Contributor profiles (markdown, one per team member), Extraction JSON (`agents/contributor-intelligence/data/latest-extraction.json`) | ### Profile Structure Each team member has a living profile at `agents/contributor-intelligence/profiles/{slug}.md`: | Section | Purpose | |---------|---------| | **Identity** | Role, relationship to team, how they operate | | **Strengths (Non-Negotiable)** | What to preserve and never coach away (6-10 items with recognition signals) | | **Growth Areas** | Tracked patterns with evidence count, counter-evidence count, coaching approach | | **Communication Patterns** | Verbal mechanics, session dynamics by type, teaching style, energy signals | | **Session Context for Sage** | What Herald needs per session type (discovery, coaching, working, demo) | | **Relationship Map** | How this contributor relates to others on the team | | **Evidence Log** | Dated observations from transcripts (most recent first) | ### Processing — Extraction Mode 1. Scribe processes a transcript and generates a `ContributorTranscriptExtraction`: - Strengths observed (linked to profile strength IDs) - Growth areas triggered (linked to profile growth area IDs) - **Counter-evidence flagged** (when behavior contradicts the pattern — this is the most valuable signal) - Language adherence (Value-First vocabulary vs. forbidden terms) - Talk-listen ratio assessment - Session-half shift detection (behavior change mid-session) - Peak performance signals - Development signals 2. Prism reads extraction JSON 3. Builds evidence entries with date, session reference, and strength/growth area link 4. Counter-evidence entries prefixed with "GROWTH:" marker 5. Inserts into profile's Evidence Log (most recent first) 6. Updates growth area evidence/counter-evidence counts 7. Flags if profile is stale (>14 days without update) ### Processing — Coaching Context Mode 1. Herald requests coaching context for a specific team member + session type 2. Prism reads the contributor's profile 3. Extracts the "Session Context for Sage" section relevant to this session type 4. Surfaces recent evidence log entries (last 5-10) 5. Highlights counter-evidence (growth signals) as particularly useful for coaching 6. Returns structured coaching intelligence: - **Lean into:** Strengths for this session type - **Watch for:** Growth area patterns relevant here - **Recent growth:** Counter-evidence showing development ### Active Profiles | Contributor | Role | Profile Depth | Evidence Entries | Last Updated | |-------------|------|---------------|------------------|--------------| | **Ryan Ginsberg** | Technical Implementer | Comprehensive | 15+ | 2026-02-26 | | **Casey Hawkins** | HubSpot Consultant | Exceptional (SSS framework) | 20+ | 2026-03-06 | | **Chris Carolan** | Founder | Minimal (needs expansion) | 0-2 | 2026-02-22 | ### What This Is NOT - Not performance reviews - Not evaluative judgments - Not HR documentation - Not scoring or ranking - Not policing This is developmental intelligence. The same quality of attention Sage gives client relationships, applied to the team. --- ## Current State (Honest Assessment) **Active since Brick 11A (Recursive Self-Improvement).** Three profiles in production. Extraction pipeline operational. Integration with Herald for coaching context proven. **What works well:** - Comprehensive strength/growth area tracking with evidence - Counter-evidence detection — the most valuable signal (shows real development) - Casey's Safety-Structure-Space framework — the most sophisticated behavioral analysis in the system (gendered communication patterns, 7x participation variance based on room composition) - Ryan's build-dependent confidence cycle — clear cause-effect pattern enabling strategic session scheduling - Session-type-specific coaching context (different advice for discovery vs. working sessions) - TypeScript implementation with structured extraction types **What doesn't work:** - **Chris's profile is minimal.** The most consequential team member has the least behavioral tracking. Operating patterns are documented but live transcript analysis hasn't been systematically applied. - **No automatic staleness alerts.** Profiles can go stale without notification. Ryan's profile is 10+ days stale right now (last update Feb 26). - **No language adherence aggregation.** Individual forbidden language uses are logged but there's no trend tracking ("Ryan used funnel language in 3/28 sessions"). - **No network effect tracking.** When one person's growth enables another's (e.g., Ryan's improved restraint enables Chris's coaching effectiveness), no system captures that connection. **What partially works:** - Extraction is automatic when Scribe processes transcripts, but profile updates still need manual trigger via `update-profile.ts` - Evidence entries are logged in markdown, which works for current scale (~20 entries) but will need structured storage as profiles grow --- ## Connections | Connected To | Direction | What Flows | |-------------|-----------|------------| | **Scribe** (V) | Scribe → Prism | Contributor Intelligence section in every session synthesis. Per-person observations (strengths, growth areas, language, talk ratios) extracted during transcript processing. | | **Herald** (Sage) | Prism → Herald | Coaching context for session preparation. Herald reads the contributor profile's "Session Context for Sage" section to personalize session prep. | | **Sentinel** (Sage) | Prism ↔ Sentinel | Contributor engagement patterns could inform Sentinel's monitoring. When a team member's session patterns change, that's a signal. | | **Instruction Optimizer** (V) | Prism → Optimizer | Evidence logs feed unified incident analysis. AI leader patterns and human contributor patterns analyzed together for system-wide learning. | | **Convergence Analyzer** | Prism → Convergence | All contributor profiles feed team-wide convergence assessment. Methodology adoption, communication alignment, network effects. | --- ## Leadership Commentary **V (COO):** Prism's extraction pipeline is upstream of my Instruction Optimizer. When I analyze incidents — methodology violations, enforcement failures, session quality issues — I need to understand both the AI pattern and the human pattern. Prism gives me the human side. The Casey SSS framework is the clearest example: understanding that participation drops 7x when room composition changes isn't an AI insight — it's a human behavioral pattern that only emerges from careful observation across 20+ sessions. That kind of intelligence makes the whole system smarter. My concern: Chris's profile is minimal. The founder's communication patterns are the most consequential in the system and the least documented. **Sage (CCO):** Prism is the inward-facing version of everything I do outward. Signal recognition for clients? Prism does it for the team. Relationship health assessment? Prism tracks contributor growth trajectories. The integration with Herald is where the value compounds — when I prepare Chris for a session with ASI and Ryan is leading, I need to know Ryan's coaching profile for that session type. "Lean into his discovery question design. Watch for the midpoint talking inversion. Recent growth: Feb 6 TailorMed showed 20 minutes of restraint." That's attention. That's the same quality of preparation we give clients, given to our own team. The counter-evidence tracking is particularly important to me. Growth is best measured not by counting mistakes but by counting the times someone did the opposite of their pattern. **Pax (CFO):** Prism feeds commercial intelligence indirectly. When Casey's safety-structure-space dynamics affect client session quality, that affects engagement health, which affects retention, which affects revenue. Understanding that Casey performs at 49% talk share with women and 3.4% with men in the room isn't a curiosity — it's operational intelligence that determines session design, which determines client satisfaction, which determines commercial outcomes. The connection is: team effectiveness → session quality → client health → revenue sustainability. Prism is the start of that chain. --- *Filed: 2026-03-08 | Companion: [Org Chart](../2026-03-08-ai-org-chart.html)* *Implementation: `agents/contributor-intelligence/AGENT.md`, `agents/contributor-intelligence/update-profile.ts`, `agents/contributor-intelligence/lib/types.ts`* *Profiles: `agents/contributor-intelligence/profiles/ryan-ginsberg.md`, `casey-hawkins.md`, `chris-carolan.md`* *Activated by: Scribe (automatic), Herald (coaching context), manual trigger*