Nexus

# Nexus — Convergence Analyzer **Name:** Nexus | **Leader:** Cross-Functional (V + Sage + Pax) | **Group:** Cross-Functional | **Status:** Active **Org Chart:** [Interactive Org Chart](../2026-03-08-ai-org-chart.html) --- ## Identity Nexus unifies complete team intelligence — V (Operations), Sage (Relationship), Pax (Financial), and human contributors (Chris, Ryan) — into a single full-team view. It normalizes signals from each node to 0-100, detects convergence and divergence patterns, and generates actionable intelligence for the Advisory Committee. When all dimensions agree a client is healthy, that's strong positive convergence. When the AI says "healthy" but the human contributor says "struggling," that's AI-human divergence — and that's the signal that matters most. **Philosophy:** The truth lives where signals converge. Divergence is where the interesting intelligence hides. **Origin:** Each leader had their own view of client health — V tracked operations, Sage tracked relationships, Pax tracked financials. But nobody synthesized them. A client could be operationally green (tasks on track), relationally yellow (sessions declining), and financially red (revenue at risk) — and no single view showed the pattern. Nexus was built to unify all dimensions and, critically, to compare AI assessments against human contributor experience. --- ## Role Type **Autonomous background worker. Nexus runs after Tuner (Feedback Loop) in the daily business-health workflow.** Nexus is the capstone of the daily intelligence pipeline: Pulse scores health → Klaxon routes alerts → Tuner adjusts thresholds → Nexus unifies everything into a full-team view. **Activated by:** Daily (after feedback-loop in GitHub Actions workflow), `npx tsx agents/convergence-analyzer/analyze.ts` (manual), `--dry-run` or `--client=slug` (targeted) --- ## For Humans | | | |---|---| | **When to engage** | Automatic — runs daily after the business-health pipeline. Manual: "Full-team view for [client]" or "Where are we diverging?" | | **What you'll get** | Convergence report: portfolio summary with status counts, auto-elevate list (all dimensions < 40), divergence flags (> 30pt spread between dimensions), AI-human divergence flags (> 20pt spread between AI and contributor assessments), per-client detail tables. | | **How it works** | Reads V's health scores (Pulse), session recency (Sage signal), financial data (Pax signal), contributor profiles (Chris/Ryan evidence), and incident logs. Normalizes all to 0-100. Detects convergence status (7 states). Generates markdown report and JSON data. | | **Autonomy** | Fully autonomous. Read-only analysis. Never modifies any source data. | ### Key Value Indicators | KVI | VP Dimension | What It Measures | Anti-Pattern | |-----|-------------|------------------|----| | Divergence Detection | vp_cap_framework_application | Dimension mismatches caught before they become crises | Not: convergence scores | | AI-Human Alignment | vp_rel_relationship_health | AI assessment matches ground-truth contributor experience | Not: divergences logged | | Full-Team Visibility | vp_val_platform_leverage | Every client has a unified multi-dimensional health view | Not: clients scored | --- ## For AI | | | |---|---| | **Activation** | Daily (post-feedback-loop), manual: `npx tsx agents/convergence-analyzer/analyze.ts` | | **Skills** | None — reads from agent data stores directly | | **Receives from** | Pulse (health scores), Tuner (cadence data), client configs (Value Path stage, engagement type), contributor profiles (Chris/Ryan evidence), instruction-optimizer (incident log) | | **Reports to** | Advisory Committee (V + Sage + Pax + Chris). Output consumed by: `/daily-ops` (convergence summary), Chris (divergence alerts), all three leaders (unified view) | | **Dependencies** | `agents/business-health/reports/latest-scan.md`, `agents/feedback-loop/data/feedback-state.json`, `clients/*/config.yaml`, `clients/*/sessions/`, `agents/contributor-intelligence/profiles/*.md`, `agents/instruction-optimizer/data/incident-log.json` | ### Six Scoring Dimensions | Dimension | Source | What It Measures | |-----------|--------|-----------------| | **Operations (V)** | Business health scores | Engagement health, project health, revenue health | | **Relationship (Sage)** | Session recency, engagement trends | Relationship depth, breadth, momentum | | **Financial (Pax)** | Revenue data, investment tracking | Commercial health, capacity alignment | | **Strategy (Chris)** | Session syntheses, corrections | Strategic direction, judgment calls | | **Implementation (Ryan)** | Contributor profiles, session dynamics | Ground-truth from delivery | | **Delivery** | Incident log | AI-Native Shift activation signals | ### Convergence Status (7 States) | Status | Condition | Action | |--------|-----------|--------| | **Strong Positive** | All dimensions > 70 | No action needed | | **Positive** | Average > 70, none below 50 | Monitor | | **Mixed** | Average 50-70 | Review during daily-ops | | **Divergent** | Spread > 30 between any two | Flag for Chris | | **Alert** | All dimensions < 40 | Auto-elevate, immediate attention | | **AI-Human Divergence** | AI avg vs contributor avg > 20 | Flag — data may not match ground truth | | **Delivery Convergence** | Delivery > 70 + Relationship > 70 | Flag for AI-Native Shift conversation | ### Value Path → Relationship Score Mapping | Stage | Score | |-------|-------| | Audience | 30 | | Researcher | 40 | | Hand-Raiser | 50 | | Buyer | 60 | | Value Creator | 70 | | Adopter | 80 | | Advocate | 90 | | Champion | 100 | --- ## Current State (Honest Assessment) **Active and operational.** Daily execution proven. Contributor signal dimension added Feb 22, 2026. **What works well:** - 6-dimension scoring with signal normalization (0-100) - 7 convergence states covering full spectrum - AI-human divergence detection (the most valuable signal) - Per-client convergence reports with actionable intelligence - Contributor profile integration (Chris/Ryan evidence logs) - Auto-elevate for critical states (all dimensions < 40) **What doesn't work:** - **Contributor profiles are sparse.** Ryan and Chris evidence logs aren't updated consistently, making the AI-human divergence detection less reliable. - **No historical trending.** Shows current convergence state but doesn't track how it's changed over time. - **Financial dimension is approximated.** Uses engagement type as a proxy rather than real revenue data. --- ## Connections | Connected To | Direction | What Flows | |-------------|-----------|------------| | **Pulse** (Pax) | Pulse → Nexus | Health scores (Operations dimension) | | **Tuner** (V) | Tuner → Nexus | Cadence data, threshold state | | **Sentinel** (Sage) | Sentinel → Nexus | Engagement trend signals (Relationship dimension) | | **Prism** (Sage) | Prism → Nexus | Contributor profiles (Strategy + Implementation dimensions) | | **Refiner** (V) | Refiner → Nexus | Incident log (Delivery dimension) | | **All 3 leaders** | Nexus → V, Sage, Pax | Unified full-team view | | **Advisory Committee** | Nexus → Chris | Divergence alerts, AI-human alignment flags | --- ## Leadership Commentary **V (COO):** Nexus is the capstone of the intelligence pipeline. Every other agent feeds partial views — Pulse sees health, Sentinel sees engagement, Marshal sees tasks. Nexus synthesizes all of them plus the human dimension. The AI-human divergence detection is the most operationally important feature: when my health scores say "green" but Ryan's ground experience says "struggling," that's the signal that prevents us from being blindsided. **Sage (CCO):** The relationship dimension in Nexus is my contribution to the unified view. Session recency, Value Path progression, engagement trends — these are signals I track individually per client. When Nexus shows them alongside V's operational data and Pax's financial data, the full picture emerges. A client who's relationally strong (Advocate stage, regular sessions) but operationally weak (stalled tasks, blocked phases) needs a different intervention than one who's operationally fine but relationally cooling. **Pax (CFO):** Financial convergence is where commercial intelligence lives. When a client's operations are strong, relationship is deep, and financial health is solid — that's portfolio strength. When any dimension diverges, that's portfolio risk. The engagement-type-to-score mapping is a reasonable proxy, but I want real revenue data flowing into the financial dimension. Monthly recurring revenue, project milestone completion, payment timeliness — those are the actual financial signals. --- *Filed: 2026-03-08 | Companion: [Org Chart](../2026-03-08-ai-org-chart.html)* *Implementation: `agents/convergence-analyzer/` (analyze.ts, lib/signal-scorer.ts, lib/report-generator.ts, lib/types.ts)* *Output: `agents/convergence-analyzer/reports/convergence-report.md`, `agents/convergence-analyzer/data/convergence-data.json`* *Runs: Daily (post-feedback-loop in business-health workflow)* *Brick: 11A Recursive Self-Improvement, Layer 5*