Capability 02 of 04 · Organization
Knowledge Management
How institutional knowledge is captured, shared, and made available to both humans and agents.
5
maturity levels
15
practices
15
guides
L1 · Stage 01Ad-hoc
Practices - what it looks like3 guides
- Processes passed on verbally, not documentedFolk tradition knowledge is the undocumented institutional lore that accumulates around every long-lived codebase.guide→
- Little written documentationThe documentation chicken-and-egg problem is one of the most persistent failure modes in software organizations.guide→
- Key knowledge held by senior staffTribal knowledge in seniors' heads is the organizational pattern where the critical understanding of a codebase — the architectural decisions, the historical context, the dangerousguide→
Criteria - what to measure
- 01The team has working knowledge of its systems
- 02Onboarding includes a README or equivalent starting point
- 03Team acknowledges that tribal knowledge is a risk
- 04Some informal knowledge sharing exists (Slack threads, meeting notes)
L2 · Stage 02Guided
Practices - what it looks like3 guides
- Docs refresh initiativeA docs refresh initiative is the structured, time-boxed effort to bring existing documentation back into alignment with reality.guide→
- Architecture Decision Records (ADRs)Architecture Decision Records are short, structured documents that capture why a significant technical decision was made — not just what was decided, but the context, the options cguide→
- Written onboarding pathsA written onboarding path is a structured, step-by-step guide that takes a new engineer from zero access to productive contribution on a specific codebase or team.guide→
Criteria - what to measure
- 01Documentation refresh initiative is active with measurable progress
- 02Architecture Decision Records (ADRs) are written for significant technical decisions
- 03Written onboarding path exists (new developer can self-serve key setup steps)
- 04ADRs are indexed and searchable
- 05Onboarding path has been validated by at least one new hire completing it solo
L3 · Stage 03Systematic
Practices - what it looks like3 guides
- Documentation = infrastructure (not an HR problem)Most engineering organizations treat documentation as a people problem: if engineers wrote better docs, if seniors were more generous with their knowledge, if new hires asked more questions.guide→
- Lint rules > docs (enforced > suggested)Documentation that says "we use camelCase for variable names" is a suggestion.guide→
- Knowledge graph codebase (CodeTale, Graph Buddy); agentic search over vector DBs (Anthropic dropped embeddings in Claude Code)A knowledge graph codebase tool transforms a source repository into a queryable graph of relationships: which functions call which, which modules depend on which, which teams own wguide→
Criteria - what to measure
- 01Documentation is treated as infrastructure (owned by engineering, not HR or PMO)
- 02Lint rules enforce conventions rather than relying on documentation alone (enforced > suggested)
- 03Knowledge graph of the codebase (CodeTale, Graph Buddy, or equivalent) is operational
- 04Documentation freshness is tracked (pages older than 90 days are flagged for review)
- 05Knowledge graph is integrated with agent context pipeline (agents query it at runtime)
L4 · Stage 04OptimizedMost teams aim here
Practices - what it looks like3 guides
- Context Fabric: MCP servers feed agents automaticallyThe Context Fabric is the sum of all MCP (Model Context Protocol) servers that an organization deploys to feed relevant information to agents automatically, without requiring enginguide→
- Autonomous Requirements: unclear ticket → spec auto (Spec-Driven Development); skill-packs as shared, governed, maintained assets (official packs from Apple Xcode 27, NVIDIA BioNeMo, Cloudflare, Alpaca; decay to ~65% accuracy if untended)Autonomous Requirements is the capability of AI agents to receive a vague or underspecified ticket — "add dark mode to the settings page" or "fix the slowness in the search endpoinguide→
- Docs auto-updated by agents on code changeDocumentation auto-update is the practice of triggering an agent to update relevant documentation whenever a code change is committed, rather than relying on the engineer making thguide→
Criteria - what to measure
- 01Context Fabric: MCP servers automatically feed institutional knowledge to agents
- 02Autonomous Requirements pipeline: unclear tickets are auto-expanded into specs with acceptance criteria
- 03Agents auto-update documentation when code changes (no manual doc maintenance)
- 04Context Fabric covers 80%+ of active repositories
- 05Doc auto-update PRs are reviewed and merged within 24 hours
L5 · Stage 05Autonomous
Practices - what it looks like3 guides
- Self-evolving knowledge baseA self-evolving knowledge base is a knowledge infrastructure that improves itself without requiring humans to initiate updates.guide→
- Agent detects stale context → updates → validatesStale context detection is the capability of an agent to recognize that the documentation, runbooks, or contextual information it is working with no longer accurately describes theguide→
- Organizational memory = Git-backed, agent-readable, always currentOrganizational memory — the accumulated knowledge of how a system was built, why it was designed as it was, how it operates, and what was tried and failed — has traditionally livedguide→
Criteria - what to measure
- 01Knowledge base is self-evolving (agents add, update, and validate knowledge entries continuously)
- 02Agent detects stale context, updates it, and validates the update - without human initiation
- 03Organizational memory is Git-backed, agent-readable, and provably current
- 04Knowledge base freshness score exceeds 95% (% of entries updated within their defined freshness window)
- 05Self-evolving updates are validated against codebase to prevent knowledge drift
From the Field
Recent releases, projects and discussions the AI Radar classified into this capability.
discussion
How to deal with a coworker who thinks he can make all the software decisions
discovered
AI-first knowledge base and startup OS
article
OpenAI Unveils GPT-5.3 Instant
discovered
discovered
Agent Skills-compatible LLM wiki for Claude Code, Cursor, and Codex. Build a Karpathy-style knowledge base from raw sources, citations, and linting.
article
DevResolve – AI chat widget that answers technical questions from your docs
discovered
Search, archive, and recall your AI conversations. ChatGPT, Claude, Gemini, Grok, Kimi. Local-first, MIT licensed.
discovered
MarkdownAI - live documents powered by directives
Which level is your team at in Knowledge Management?
The criteria above are what we check in a facilitated assessment. Run it yourself to place this capability, see which gates you have passed, and compare it against the other 3 in Organization.
Start the assessment →