Portfolio Operating Model
How the ecosystem operates — the complete lifecycle of work, from information entering the ecosystem to customer outcomes and feedback. Where the architecture pages describe what exists and the capabilities describe what is reusable, this section describes how the ecosystem functions, evolves, and governs itself. Evidence-first; uncertainty recorded.
This is the Operating Model section. Its pages:
| Page | What it covers |
|---|---|
| This page | The end-to-end operating lifecycle of work |
| Knowledge Asset Lifecycle | Raw sources → … → commercial outcomes, per stage |
| Concept Ownership Registry | Who owns each architectural concept; ambiguous ownership flagged |
| Portfolio Glossary | The authoritative, de-duplicated vocabulary |
| Context Architecture | Repository → capability → platform → portfolio → business context |
| Operating Principles | How work is performed (vs how systems are structured) |
| AI Collaboration Model | How humans, AI assistants, orchestration & future agents collaborate |
The operating lifecycle
Work flows through seven stages. This mirrors the Knowledge Asset Lifecycle but focuses on activity — who or what does the work at each step.
graph LR
A["① Information enters<br/>knowledge estate · captures · signals · website URLs"] --> B["② Processed<br/>harvest/extract → stage → structure"]
B --> C["③ Intelligence created<br/>scoped, maturity-gated knowledge assets"]
C --> D["④ Products<br/>versioned immutable packs"]
D --> E["⑤ Ventures consume<br/>apply packs · assess · qualify"]
E --> F["⑥ Customer outcomes<br/>reports · proposals · outreach · journeys"]
F --> G["⑦ Feedback<br/>outcomes → (future) re-weight intelligence"]
G -. one-way, new pack version .-> D
Stage-by-stage (evidence-based)
| Stage | What happens | Primary systems | Evidence |
|---|---|---|---|
| ① Enters | Founder knowledge, ad-hoc captures (voice/articles/agent research/legacy KB), market signals, and website URLs enter | FIP; the assessment engine | Knowledge Intelligence, Website Assessment |
| ② Processed | Extraction/harvest → staging queues → human review & approval → structuring | FIP (PersonalOps) | personalops digest |
| ③ Intelligence | Approved intelligence becomes scoped, maturity-gated knowledge_assets (working→ready→proven) |
FIP | personalops digest |
| ④ Products | Builder skills synthesise assets into versioned, immutable packs | intelproducts (published by FIP) | Intelligence Productization |
| ⑤ Ventures consume | Engine applies a pinned pack to a website (crawl→score→findings→recs); apps qualify/segment | inexisstudios, outreachagent, leadplatform | Website Assessment |
| ⑥ Customer outcomes | Reports, COPs, proposals, outreach, journeys reach customers | ventures (Inexis Digital, Inbound Lanka, …) | cap-commercial-opportunity |
| ⑦ Feedback | Outcomes signal quality; planned Hermes loop re-weights intelligence into a new pack version | (future) Hermes | Intelligence Platform |
Uncertainty (recorded): stage ⑦ is not built (Hermes is a reserved future loop), and stage ⑤ cross-repo consumption is mostly mock in
outreachagent. The loop is designed, not closed.
Activity types across the lifecycle
The same lifecycle, coloured by who or what performs the work — the distinction the ecosystem must be explicit about as it automates.
graph TD
subgraph HUMAN["👤 Human activities"]
H1[Capture intelligence]
H2[Review & approve knowledge assets]
H3[Approve assessment report]
H4[Approve outreach send — 'nothing auto-sent']
H5[Architectural governance / ADRs]
end
subgraph AI["🤝 AI-assisted activities"]
AI1[Harvest / extraction scripts]
AI2[AI explains findings — narrative only]
AI3[Skill invocations - Claude Code / ChatGPT]
AI4[Pack authoring assistance]
end
subgraph AUTO["⚙️ Automated - deterministic"]
AU1[Rule-based scoring 0–100]
AU2[Deterministic retrieval - search_assets, no embeddings]
AU3[Immutable pack versioning]
AU4[Website crawl]
end
subgraph FUTURE["🔮 Future autonomous"]
F1[OpenClaw agent - governed service consumer]
F2[Hermes learning loop]
F3[portfolio-portal-orchestrator - self-documenting]
F4[Agent-initiated harvests]
end
H1 --> AI1 --> H2 --> AU2
AU1 --> AI2 --> H3
AI3 -.-> AU3
F1 & F2 & F3 & F4 -.planned.-> H5
classDef future fill:#eee,stroke-dasharray:4 3;
class F1,F2,F3,F4 future;
| Activity type | Definition | Examples (evidenced) | Status |
|---|---|---|---|
| 👤 Human | A person performs or must approve the step | Capture; mandatory approval gates (FIP asset promotion, report approval, outreach send); architectural governance | ✅ Implemented (the human boundary is by design, permanent for MVP) |
| 🤝 AI-assisted | AI does the work under human direction/skills; a human reviews | Harvest/extraction scripts; AI explains findings (never decides); Claude Code / ChatGPT via skills & handoff files | ✅ Implemented |
| ⚙️ Automated (deterministic) | Runs without a model; exact & repeatable | Rule-based scoring; deterministic retrieval (no embeddings at runtime); immutable pack versioning; crawl | ✅ Implemented |
| 🔮 Future autonomous | Agents act with limited human oversight | OpenClaw integration; Hermes learning; orchestrator auto-documentation; agent-initiated harvests | ⏳ Planned / not built |
The governing rule (evidenced): across the ecosystem, deterministic rules decide, AI explains, and a human approves before anything external happens (Principle 11; Operating Principles). Automation increases within stages; the human approval boundary (capture, review, approval) is deliberately preserved, not a temporary constraint.
How the operating model governs itself
See the dedicated Architecture Governance section for how change is designed, reviewed, approved, and incorporated (decision framework, repository lifecycle, AI governance, review lifecycle). In brief:
- Decisions are recorded as ADRs; direction changes are proposed, not silently applied.
- Documentation is evidence-first and honestly labelled (Principle 8); this portal is the operating manual.
- Change is phase-gated with architecture review (Principle 10).
- Reuse & risk are tracked in the Capability Reuse Map and Architecture Risks.