AIFC-010: Knowledge Structure

Status: Draft 0.1 Standard: AI-First Community Standard Short name: AIFC Related to:

• AIFC-000: Manifesto for AI-First Communities
• AIFC-001: Core Concepts
• AIFC-002: Community Model
• AIFC-003: Values and Purpose
• AIFC-004: Feedback and Change Proposals

Purpose of this document: Define the basic structure of an AIFC community knowledge base: which knowledge types it should contain, how they should be organized, how they should serve people, AI agents, and validators, and how ordinary documentation becomes the community’s Source of Truth.

1. Purpose of this document

This document describes the knowledge structure of an AIFC community.

An AIFC community needs a Source of Truth that is not only an archive of documents, but a living knowledge structure for:

• purpose,
• values,
• strategy,
• decisions,
• workflows,
• skills,
• work,
• feedback,
• AI governance,
• security,
• Community Interfaces,
• and system learning.

The knowledge base should be usable in three ways:

human-readable
agent-actionable
software-verifiable

A person must be able to understand it. An AI agent must be able to use it. Software must be able to validate it.

2. Core principle

The core principle of this document is:

Knowledge base is not an archive.
Knowledge base is the living memory and operational structure of the community.

An AIFC knowledge base must not be only a place where texts are stored.

It must help the community:

• understand itself,
• hold purpose,
• decide,
• work,
• learn,
• protect its know-how,
• involve AI safely,
• and remain operable without AI.

3. Knowledge base vs documentation

AIFC distinguishes ordinary documentation from a knowledge base.

Documentation

Ordinary documentation often describes:

• what exists,
• how something works,
• how to use something,
• what happened,
• what was decided.

It may be useful, but it is often:

• fragmented,
• outdated,
• without ownership,
• disconnected from decisions,
• without clear structure,
• difficult for AI to use,
• difficult for software to validate.

Knowledge base

An AIFC knowledge base is structured community memory.

It does not only describe the past. It supports present and future community operation.

It must make it possible to answer:

• Why does the community exist?
• Which values does it hold?
• What is the current state?
• What is the desired state?
• What path connects them?
• Which decisions were made?
• Which work is in progress?
• Which skills do people and AI use?
• Which AI rules apply?
• Which risks exist?
• What is waiting for change?
• What has the community learned?
• How does it cooperate with other communities?

Minimum requirement

An AIFC knowledge base must be more than a collection of documents.

It must contain structured artefacts that support purpose, decisions, work, learning, and governance.

4. Source of Truth

The knowledge base must have a defined Source of Truth.

The Source of Truth is the authoritative place where the community keeps current and approved knowledge.

The Source of Truth may be implemented in different ways, but AIFC prefers a format that is:

• open,
• versionable,
• exportable,
• human-readable,
• agent-actionable,
• software-verifiable.

Markdown with metadata in Git is a natural candidate because it satisfies many of these requirements.

AIFC is not dependent on one specific technology. The principle matters.

Minimum requirement

The community must know:

• where the authoritative knowledge base is,
• who owns it,
• how it changes,
• who approves changes,
• how changes are audited,
• how it is backed up,
• how it is exported,
• how it can be used without AI.

5. Human-readable, agent-actionable, software-verifiable

Every key knowledge artefact should be designed with three layers.

5.1 Human-readable

A person must understand the meaning.

The artefact should be:

• clear,
• concise,
• contextual,
• readable without a special tool,
• understandable to a new community member,
• usable in decision-making.

5.2 Agent-actionable

An AI agent must be able to use the artefact for work.

The artefact should have:

• clear purpose,
• structured inputs,
• expected outputs,
• rules,
• constraints,
• links to other artefacts,
• information about what the agent may and must not do.

5.3 Software-verifiable

Software must be able to verify at least some rules.

For example:

• missing owner,
• missing status,
• missing Decision Record,
• AI workflow without fallback,
• restricted data without an AI-NDA Boundary,
• skill without a source,
• change proposal without a decision level.

Minimum requirement

Critical knowledge artefacts must be designed so they are not only text, but have verifiable structure.

6. Human Cockpit Layer

The Source of Truth is the community’s memory.

The Human Cockpit Layer is human access to that memory.

The knowledge base may be well structured for AI agents, validators, and audit, but too complex for ordinary community members if it is available only as a file tree, metadata, and Git history.

AIFC therefore distinguishes:

Source of Truth
-> authoritative structured knowledge

Human Cockpit Layer
-> human-operable layer over that knowledge

The Human Cockpit Layer should help people:

• orient themselves,
• read the current state,
• see purpose and values,
• understand priorities,
• approve changes,
• follow feedback,
• work with change proposals,
• see AI involvement,
• recognize risks,
• understand relationships.

Minimum requirement

An AIFC community must have a human-accessible way to work with its knowledge base.

It does not have to be a standalone application, but an ordinary community member must not be limited only to the repository’s technical structure when that structure exceeds their ability to orient themselves.

7. Minimal knowledge domains

An AIFC knowledge base should contain at least these domains:

purpose
values
strategy
current-state
desired-state
path
decisions
principles
workflows
skills
work
feedback
risks
ai-governance
security
interfaces
retrospectives

These domains may be implemented as folders, files, database entities, a knowledge graph, or another structure.

AIFC does not require a specific technology, but it requires these domains to be findable and usable.

8. Purpose knowledge

Purpose knowledge describes why the community exists.

It contains:

• community purpose,
• reason for existence,
• who or what the community serves,
• what value it creates,
• which desired state it supports,
• boundaries of purpose,
• history of significant purpose changes.

Minimum requirement

Purpose must be:

• explicitly described,
• approved by the community or responsible owner,
• stored in the Source of Truth,
• accessible to people,
• available to relevant AI agents,
• connected to values, strategy, and work.

9. Values knowledge

Values knowledge describes the values the community does not want to sacrifice.

It contains:

• values,
• their interpretation,
• examples of correct behavior,
• anti-patterns,
• values conflicts,
• decisions that clarified values,
• connection to AI governance,
• connection to the Community Interface.

Minimum requirement

Values must be usable in decision-making.

A value that does not affect decisions is not governance. It is decoration.

10. Strategy knowledge

Strategy knowledge describes the path from current state to desired state.

It contains:

current state
desired state
path
priorities
trade-offs
constraints
risks
assumptions
success indicators

Strategy is not only a list of initiatives.

It is the community’s conscious path between where it is and where it wants to go.

Minimum requirement

Strategy must be connected to:

• purpose,
• values,
• current state,
• desired state,
• work,
• decisions,
• feedback.

11. Decision knowledge

Decision knowledge preserves the memory of decisions.

It contains:

• Decision Records,
• rejected proposals,
• accepted proposals,
• reasons for decisions,
• affected values,
• alternatives,
• expected impacts,
• review points.

Without decision knowledge, the community repeats the same discussions and loses the ability to learn.

Minimum requirement

Significant decisions must have a Decision Record.

Each Decision Record must be traceable from related work, strategy, values, or change proposal.

12. Workflow knowledge

Workflow knowledge describes how the community performs repeatable work.

It contains:

• workflows,
• inputs,
• outputs,
• owners,
• rules,
• exceptions,
• approval gates,
• AI steps,
• non-AI fallback,
• risks,
• quality measurement.

A workflow may be:

• human-only,
• AI-assisted,
• agentic,
• hybrid,
• or fallback.

Minimum requirement

Critical workflows must have:

• owner,
• input and output,
• decision rules,
• fallback,
• AI involvement rules,
• auditability.

13. Skill knowledge

Skill knowledge describes how the community does something well.

AIFC distinguishes:

Human Skill
AI Skill

Human Skill

A Human Skill is intended for a person.

It contains:

• principles,
• procedures,
• checklists,
• examples,
• anti-patterns,
• onboarding,
• output quality.

AI Skill

An AI Skill is intended for an agent.

It contains:

• role,
• instructions,
• inputs,
• outputs,
• allowed actions,
• forbidden actions,
• examples,
• constraints,
• approval rules.

Minimum requirement

Critical skills must exist in a form usable by people.

AI skills must not be the only place where community know-how is stored.

14. Work knowledge

Work knowledge describes community work.

It contains:

• backlog,
• tasks,
• epics,
• maintenance items,
• support items,
• change work,
• owners,
• priority,
• status,
• relationship to purpose,
• relationship to strategy,
• relationship to decisions.

AIFC distinguishes at least:

• development / change work,
• maintenance work,
• support work,
• learning work,
• governance work,
• feedback processing.

Minimum requirement

Significant work must be traceably connected to purpose, strategy, values, or decisions.

Maintenance work is not second-class work. Everything the community does not care for tends to degrade or create debt.

15. Feedback knowledge

Feedback knowledge preserves signals and change proposals.

It contains:

• observed signals,
• change proposals,
• proposal status,
• proposal classification,
• decision level,
• Decision Record,
• implementation status,
• verification status.

Feedback must not remain only in conversations, meetings, or chats.

If feedback shows significant risk or opportunity, it must have a path into the Source of Truth.

Minimum requirement

Significant change proposals must be recorded, evaluated, and decided.

16. Risk knowledge

Risk knowledge describes community risks.

It contains:

• operational risks,
• security risks,
• AI dependency risks,
• AI lock-in risks,
• knowledge leakage risks,
• compliance risks,
• values conflicts,
• purpose drift,
• Ghost AI Company risk,
• impacts on other communities.

Minimum requirement

Critical risks must have an owner, status, mitigation, and link to decisions or work.

17. AI governance knowledge

AI governance knowledge describes how the community uses AI.

It contains:

• AI usage inventory,
• AI-NDA Boundaries,
• AI agents,
• AI tools,
• AI workflows,
• AI autonomy levels,
• AI operating modes,
• AI capacity planning,
• AI budget,
• human approval rules,
• fallback modes,
• AI retrospective,
• exit strategy.

Minimum requirement

The community must know:

• where it uses AI,
• which data AI sees,
• who the owner is,
• what the fallback is,
• how benefit is measured,
• how lock-in is prevented.

18. Security knowledge

Security knowledge describes protection of the knowledge base and AI involvement.

It contains:

• data classification,
• access control,
• agent permissions,
• audit rules,
• AI-NDA Boundary,
• secrets handling,
• export rules,
• incident response,
• backup,
• recovery,
• vendor risk.

AIFC recommends this minimal classification:

Public
Internal
Restricted
Operational DNA

Minimum requirement

Operational DNA must be protected as a critical asset.

An AI agent must not have access to everything only because it is useful.

19. Interface knowledge

Interface knowledge describes how the community communicates with other communities.

It contains:

• who we are,
• what we offer,
• what we need,
• which values we have,
• which boundaries we have,
• which inputs we accept,
• which outputs we provide,
• how we escalate conflict,
• how we accept change proposals,
• how we share knowledge,
• how we protect sensitive information.

Minimum requirement

An AIFC community must be able to clearly describe its interface toward other communities.

20. Retrospective knowledge

Retrospective knowledge preserves learning from past cycles.

It contains:

• sprint retrospective,
• AI retrospective,
• incident retrospective,
• project retrospective,
• lessons learned,
• AI value,
• AI waste,
• AI dependency,
• workflow conversion candidates,
• skill update candidates,
• maintenance needs,
• change proposals.

A retrospective is not only a meeting.

It is a mechanism for converting experience into system improvement.

Minimum requirement

Significant lessons learned must be converted into the Source of Truth, skills, workflows, backlog, or change proposals.

21. Metadata

Metadata prevents the knowledge base from being only text.

Metadata may describe, for example:

• type,
• status,
• owner,
• priority,
• sensitivity,
• related values,
• related purpose,
• related decisions,
• AI access level,
• review date,
• source,
• confidence,
• lifecycle,
• affected communities.

Why it matters

Metadata helps:

• navigation,
• validation,
• AI agents,
• Human Cockpit Layer,
• auditability,
• cleanup,
• automation,
• search,
• access management.

Minimum requirement

Critical artefacts must have minimal metadata:

• type,
• owner,
• status,
• sensitivity,
• last update or review information.

22. Lifecycle of knowledge

Knowledge artefacts have a lifecycle.

Recommended statuses:

draft
proposed
under_review
approved
active
deprecated
archived
rejected

Why it matters

Without lifecycle, the community does not know what to trust.

Outdated documentation creates knowledge debt. Unmaintained knowledge degrades decisions, workflows, AI skills, and human capability.

Minimum requirement

Critical artefacts must have a status and owner.

Deprecated artefacts must not be used by AI agents as the current Source of Truth.

23. Ownership

Every critical knowledge artefact must have an owner.

The owner is responsible for:

• correctness,
• currentness,
• approval of changes,
• links to related artefacts,
• review,
• sensitivity,
• AI use,
• archiving.

Why it matters

Knowledge without an owner tends to degrade.

If it is unclear who cares for knowledge, it gradually turns into debt.

Minimum requirement

Critical artefacts without an owner must be marked as a governance issue or maintenance need.

24. Review and maintenance

The knowledge base requires care.

Maintenance is not extra work. It protects the system from degradation.

An AIFC knowledge base should have regular review according to sensitivity and importance.

Examples:

• values and purpose: periodically or after a significant signal,
• strategy: regularly according to the community cycle,
• workflow: after a process change or incident,
• AI workflow: after a model, vendor, or risk change,
• skills: after significant outputs or retrospective,
• security rules: according to risk management,
• interface: after a cooperation or impact change.

Minimum requirement

Critical knowledge artefacts must have a review mechanism.

25. Knowledge capture from AI

AI often generates outputs that contain new know-how.

For example:

• better formulation of a rule,
• workflow proposal,
• skill proposal,
• risk analysis,
• design pattern,
• decision alternatives,
• lessons learned,
• new terms.

If such know-how remains only in AI chat or a proprietary tool, the community does not fully own it.

Minimum requirement

Significant know-how created with AI assistance must be evaluated for inclusion in the Source of Truth.

AI must not become the uncontrolled external memory of the community.

26. Knowledge import and migration

An AIFC community may emerge by migrating from existing sources:

• Confluence,
• SharePoint,
• Jira,
• Git,
• emails,
• documents,
• tickets,
• internal wikis,
• conversations,
• people.

Migration is not only content transfer.

It is transformation from documentation chaos into a structured Source of Truth.

AI may help:

• extract purpose,
• find duplication,
• detect outdated content,
• propose structure,
• create metadata,
• identify gaps,
• propose skills,
• create change proposals.

Minimum requirement

Migration into an AIFC knowledge base must distinguish:

• imported content,
• AI interpretation,
• human-approved knowledge,
• uncertain or conflicting parts,
• sensitivity and access.

27. Knowledge export and portability

The knowledge base must be portable.

AIFC rejects a state where operational community know-how is locked in a proprietary tool without export.

Exportability matters for:

• audit,
• backup,
• exit strategy,
• vendor replacement,
• reference implementation,
• human review,
• long-term preservation.

Minimum requirement

Critical knowledge artefacts, skills, workflows, and governance rules must be exportable in a format readable without the original tool.

28. Knowledge quality

The knowledge base must have quality.

Quality includes:

• currentness,
• correctness,
• clarity,
• ownership,
• traceability,
• decided status,
• sensitivity,
• links,
• usability for people,
• usability for AI,
• software validation.

Quality issues

Typical issues:

• duplication,
• outdated content,
• no owner,
• no status,
• unclear decision,
• conflicting rules,
• missing fallback,
• missing AI-NDA Boundary,
• missing metadata,
• knowledge trapped in chat,
• skill not updated after learning.

Minimum requirement

An AIFC community must have a mechanism for identifying and resolving knowledge quality issues.

29. Suggested minimal folder/domain structure

This structure is illustrative, not mandatory.

/aifc-community
  /purpose
  /values
  /strategy
  /knowledge
  /decisions
  /workflows
  /skills
    /human
    /ai
  /work
    /backlog
    /maintenance
    /support
    /change
  /feedback
    /signals
    /change-proposals
  /retrospectives
  /ai-governance
  /security
  /interfaces
  /risks
  /rules

The standard does not mandate one folder tree for every implementation.

It requires the domains to be findable, consistent, and validatable.

30. Anti-patterns

AIFC rejects the following anti-patterns.

30.1 Documentation chaos

Many documents without clear structure, owners, status, and links to decisions.

30.2 Knowledge trapped in people

Critical know-how exists only in the heads of individuals.

30.3 Knowledge trapped in AI tools

Know-how created with AI remains in chat, agent memory, or a proprietary tool.

30.4 Structure without human access

The knowledge base is well structured for AI and validators, but people do not understand or use it.

30.5 Human-readable only

Documentation is understandable to people but has no structure for AI and validation.

30.6 Machine-readable only

The structure fits software, but people cannot read, approve, or maintain it.

30.7 No Source of Truth

It is unclear which knowledge is current and approved.

30.8 No ownership

Critical knowledge has no owner.

30.9 No lifecycle

It is unclear what is draft, approved, deprecated, or archived.

30.10 No review

Knowledge is not maintained and degrades.

30.11 AI reads deprecated knowledge

An AI agent uses outdated or unapproved content as current truth.

30.12 Operational DNA without protection

The community’s most valuable know-how is structured, but not adequately protected.

31. Minimal requirements

For knowledge structure, an AIFC community must at least:

1. Have a defined Source of Truth.
2. Keep the knowledge base human-readable, agent-actionable, and software-verifiable.
3. Have a human-accessible layer for working with the knowledge base.
4. Contain purpose.
5. Contain values.
6. Contain strategy or path.
7. Contain decisions or Decision Records.
8. Contain workflows for critical activities.
9. Contain human skills for critical capabilities.
10. Contain AI governance rules for significant AI use.
11. Contain feedback and a change proposal mechanism.
12. Contain security and data classification rules.
13. Give critical artefacts an owner.
14. Give critical artefacts status or lifecycle.
15. Give critical artefacts sensitivity.
16. Give critical AI workflows fallback and an AI-NDA Boundary where relevant.
17. Evaluate significant AI-generated know-how for inclusion in the Source of Truth.
18. Keep the knowledge base exportable or portable without major loss of meaning.
19. Maintain a review and maintenance mechanism.
20. Protect Operational DNA as a critical asset.

32. Summary

An AIFC knowledge base is not an archive.

It is the living memory of the community.

It contains purpose, values, strategy, decisions, work, skills, workflows, feedback, security, AI governance, and interfaces with other communities.

Its quality determines how well the community can:

• understand itself,
• manage work,
• learn,
• involve AI,
• protect know-how,
• transfer knowledge to new people,
• operate fallback without AI,
• cooperate with other communities.

The Source of Truth is memory.

The Human Cockpit Layer is human access to that memory.

AI is an accelerator over that memory.

The community remains the owner of purpose, values, and decisions.

AIFC knowledge structure turns documentation into governed community memory.