AIFC
Back to version

Published version: AIFC-V002. This is the latest published version. All versions.

AIFC-001: Core Concepts

Status: Draft 0.1 Standard: AI-First Community Standard Short name: AIFC Related to: AIFC-000: Manifesto for AI-First Communities Purpose of this document: Define the core concepts used across the AIFC standard so they are understandable to people, usable by AI agents, and verifiable by software.

1. Purpose of this document

This document defines the core concepts of the AIFC standard.

AIFC-000 explains why the standard exists and which principles it protects. AIFC-001 stabilizes the language so the rest of the standard can be precise, repeatable, and agent-actionable.

Each concept is described in four layers:

2. Definition style

AIFC uses concepts so they are:

Where useful, the standard distinguishes:

AIFC is not only a philosophical framework. It should support documentation, governance, validators, agents, workflows, skills, and reference implementations.

3. Community

Definition

Community is a group of people or represented actors that shares purpose, values, knowledge, work, decisions, or responsibility.

A community may be:

Why it matters

AIFC does not place AI, an application, or documentation at the center. It places the community at the center.

AI should serve the community. The community owns purpose, values, responsibility, and direction.

Minimum requirement

An AIFC-compatible community must have at least:

Purpose, Values, Human Ownership, Community Interface, Feedback Loop.

4. Purpose

Definition

Purpose is the reason a community exists and the direction it consciously wants to move toward.

Purpose answers questions such as:

Why it matters

AI can optimize work very quickly. If a community lacks a clear purpose, AI may only accelerate chaos.

Purpose determines what AI should accelerate.

Minimum requirement

A community must record its purpose in its Source of Truth.

Purpose must be:

Values, Strategy, Human Ownership, Community Model, Feedback Loop.

5. Values

Definition

Values are commitments the community does not want to sacrifice under pressure for speed, performance, efficiency, or automation.

Values are the highest decision layer of the community.

Why it matters

AI can quickly optimize for a badly defined goal.

Without values, an AI-first community may become a system that efficiently does the wrong things.

Minimum requirement

A community must have values that are:

The community must have a mechanism for refining the interpretation of values based on experience and feedback.

Purpose, Governance, Feedback Loop, Change Proposal, Community Interface.

6. Human Ownership

Definition

Human Ownership means that people or the community remain the owners of purpose, values, responsibility, critical decisions, and direction.

AI may propose, analyze, formulate, compare, and accelerate. AI must not be the final owner of community direction.

Why it matters

Without Human Ownership, an AI-first system may become AI-dependent or ghost-like: it produces outputs, but has no clear responsibility.

Minimum requirement

Every critical decision, AI workflow, and change of direction must have an identified human or community owner.

An AI-generated proposal must not be treated automatically as a decision.

Human-managed AI, AI as External Expert Capacity, Change Proposal, Ghost AI Company.

7. Knowledge Base

Definition

Knowledge base is the structured body of community knowledge.

It may contain:

Why it matters

The knowledge base is the community’s memory.

In an AI-first environment, the knowledge base is not only documentation. It is an input layer for people, software, and AI agents.

Minimum requirement

The knowledge base must be:

Operational DNA, Source of Truth, Metadata and Markdown, Knowledge Security.

8. Operational DNA

Definition

Operational DNA is the critical part of the knowledge base that describes how the community actually creates value, decides, learns, and functions.

It includes, for example:

Why it matters

Operational DNA is one of the most valuable assets of a community or company.

If it leaks, a competitor may gain not only documentation, but also the operating model of the community.

Minimum requirement

Operational DNA must be classified as a critical asset.

It must have:

Knowledge Security, AI-NDA Boundary, Company as a System, Company as Product.

9. Source of Truth

Definition

Source of Truth is the authoritative place where the community keeps current and approved knowledge, decisions, rules, skills, and governance.

Why it matters

If know-how remains in chat, an AI tool, personal memory, or a proprietary system, the community does not fully own it.

AI outputs must return to the Source of Truth when they contain new or changed know-how.

Minimum requirement

An AIFC community must define its Source of Truth.

Significant AI outputs, decisions, rule changes, and skill updates must be written to the Source of Truth or proposed for inclusion.

Knowledge Base, Operational DNA, Skill Evolution, AI Lock-in.

10. Human Cockpit Layer

Definition

Human Cockpit Layer is the human-operable layer over the community knowledge base.

It allows community members to understand, manage, approve, plan, and navigate purpose, values, work, decisions, risks, AI involvement, and change proposals without having to read the whole source Markdown, metadata, Git history, or validation schemas directly.

The Human Cockpit Layer does not have to be one specific application. It may be a web interface, dashboard, documentation portal, assisted editor, workflow UI, or another human-accessible layer over the Source of Truth.

Why it matters

An AIFC knowledge base should be human-readable, agent-actionable, and software-verifiable.

That does not mean every community member should work directly with the full repository, metadata, or validation rules.

AI agents and software may work directly with Markdown, metadata, schemas, and the Source of Truth structure. People need a layer that protects attention and helps them quickly understand:

Without a Human Cockpit Layer, the knowledge base may be well structured for AI and validators, but too difficult for people to operate.

AIFC therefore protects not only machine readability, but also the human ability to understand and manage the system.

Minimum requirement

An AIFC community must have a human-accessible way to read, manage, and decide over its knowledge base.

It does not have to be a standalone application, but ordinary community members must not be required to understand the whole technical structure of the repository, metadata, and validation rules.

The Human Cockpit Layer must at least allow:

Source of Truth, Knowledge Base, Human-readable Agent-actionable Software-verifiable, Feedback Loop, Change Proposal, AI Retrospective, Human Ownership.

11. Human-readable, Agent-actionable, Software-verifiable

Definition

AIFC artefacts should be simultaneously:

Why it matters

Documentation understood only by people is weakly usable for agents. A structure understood only by software is hard for people to manage. A prompt understood only by an AI tool creates lock-in risk.

AIFC requires all three properties together.

Minimum requirement

Every key standard artefact should have:

Metadata and Markdown, Agent Skills, Validation Rules, Standard as Code.

12. AI as Accelerator

Definition

AI as Accelerator means that AI accelerates understanding, work, decisions, synthesis, learning, and system creation.

AI is not the driver of the community.

Why it matters

AI-first does not mean AI governs the community. It means the community is designed so AI can safely accelerate its purpose.

Minimum requirement

AI use must be connected to purpose, values, and a human owner.

AI must not replace human ownership of direction.

Human Ownership, human-managed AI, AI as External Expert Capacity.

13. AI as External Expert Capacity

Definition

AI as External Expert Capacity is a metaphor and governance principle that manages AI similarly to external expert consulting capacity.

AI has know-how, speed, and ability to help, but it needs:

Why it matters

This metaphor returns AI to a familiar management frame.

A company would not let an external consulting firm into internal systems without a contract, NDA, scope, and responsibility. AI should not be an exception.

Minimum requirement

Every significant use of AI over non-public know-how must define:

AI-NDA Boundary, AI Lock-in, Knowledge Security, AI Capacity Planning.

14. AI-NDA Boundary

Definition

AI-NDA Boundary is the rules-defined confidentiality boundary for using AI with non-public community know-how.

It determines:

Why it matters

AI may act as external intelligence. Without a confidentiality boundary, data and know-how may be transferred outside the community without control.

Minimum requirement

AI must not process non-public or sensitive knowledge assets without a defined AI-NDA Boundary.

Data Classification, Knowledge Security, Operational DNA, Agent Permissions.

15. AI Capacity

Definition

AI Capacity is the limited operational AI capacity a community can use in a given period.

It includes:

Why it matters

AI is not an unlimited resource. Without planning, AI may burn budget, create noise, or increase dependency.

Minimum requirement

A community using AI beyond ad hoc work must have a way to track AI cost, benefit, and constraints.

For critical teams or workflows, AI capacity should be planned similarly to sprint capacity.

AI Capacity Planning, AI Budget, AI Retrospective, AI Waste Backlog.

16. AI Autonomy

Definition

AI Autonomy is the degree to which AI may act without continuous human confirmation.

Autonomy may be low, for example when AI only proposes. It may also be high, for example when AI performs approved low-risk steps within guardrails.

Why it matters

AI autonomy is not binary. It must be managed according to risk, data, budget, maturity, responsibility, and fallback.

Minimum requirement

Every AI agent or AI workflow must have a defined autonomy level and conditions under which human approval is required.

AI Intensity, AI Operating Mode, Human Approval, Agent Permissions.

17. AI Intensity

Definition

AI Intensity is the overall degree of AI involvement in a community or workflow.

It may be expressed on a scale such as 0-100 percent:

Why it matters

The same operating model may run at different AI intensity levels depending on budget, capital, risk tolerance, and governance maturity.

Minimum requirement

If a community runs AI workflows, it should be able to identify their AI intensity and change it according to the situation.

AI Autonomy, AI Operating Mode, AI Budget, Company as a System.

18. AI Operating Mode

Definition

AI Operating Mode is a named AI involvement mode that combines AI intensity, budget rules, risk limits, human approval, and fallback.

Examples:

Why it matters

A community does not have to use the same degree of AI all the time. One mode may fit stable operations, another crisis migration, and another experimentation.

Minimum requirement

Critical AI-first communities should have at least basic modes:

AI Intensity, AI Budget, AI Lock-in, Human Capability Reserve.

19. AI Budget

Definition

AI Budget is a reserved budget for AI work in a given period or area.

It may include:

Why it matters

Without a budget, AI cannot be managed as a limited resource.

AI budget may also serve as a mechanism for automatically reducing AI intensity.

Minimum requirement

Significant AI use must have tracked costs.

Critical AI workflows must have rules for what happens when the budget limit is reached.

AI Capacity, AI Operating Mode, AI Retrospective.

20. AI Lock-in

Definition

AI Lock-in is dependency on a specific AI vendor, model, agent memory, proprietary skill store, prompt workflow, or AI step such that the community cannot migrate or function without unacceptable capability loss.

AI lock-in may be:

Why it matters

AI lock-in is often less visible than classic software lock-in.

A community may gradually move its know-how, decision logic, and ability to work into an external AI environment.

Minimum requirement

No critical workflow may depend on one AI vendor, model, agent memory, or proprietary skill store without an exit strategy.

Exit Strategy, AI-NDA Boundary, Source of Truth, Human Capability Reserve.

21. Exit Strategy

Definition

Exit Strategy is a plan for ending or replacing cooperation with a specific AI vendor, model, agent, or tool without unacceptable impact on the community.

Why it matters

AI should be terminable.

If a community cannot turn off or replace an AI step without workflow collapse, it has an operational risk.

Minimum requirement

Critical AI workflows must have:

AI Lock-in, Human Capability Reserve, AI Operating Mode.

22. Human Capability Reserve

Definition

Human Capability Reserve is the consciously maintained human ability to perform, understand, review, or restore work without AI.

It may include:

Why it matters

If a token outage stops simple routine work, the company has not gained intelligence. It has lost resilience.

AI should accelerate community capability, not replace it so completely that the community becomes impaired without AI.

Minimum requirement

The community must preserve a non-AI path for critical capabilities.

A recommended rule is to regularly perform a defined portion of work, for example 10 percent, without AI across task types.

AI Dependency, Exit Strategy, Human Skills, AI Backup Recovery.

23. AI Dependency

Definition

AI Dependency is the state in which a community or its members cannot perform work, make decisions, or review results at acceptable quality without AI.

Why it matters

AI acceleration is healthy. AI dependency is risky.

The difference:

Minimum requirement

AI retrospective must track signs of AI dependency.

Critical AI dependencies must be managed through fallback, training, Human Capability Reserve, or workflow redesign.

Human Capability Reserve, AI Lock-in, AI Retrospective.

24. AI Retrospective

Definition

AI Retrospective is a regular evaluation of benefit, cost, risk, noise, dependency, and learning created by AI use.

Why it matters

AI retrospective converts AI consumption into organizational learning.

It helps distinguish:

Minimum requirement

A community that uses AI significantly must regularly evaluate AI usage, AI value, AI waste, AI dependency, and skill updates.

AI Waste Backlog, Workflow Conversion, Skill Evolution, AI Capacity.

25. AI Waste Backlog

Definition

AI Waste Backlog is a record of repeated AI activities that consume capacity but should be converted into a non-AI workflow, template, validator, rule, or UI function.

Why it matters

AI should not repeat routine work indefinitely if that routine can be converted into the system.

An AI waste backlog helps reduce cost, noise, and dependency.

Minimum requirement

AI retrospective should identify candidates for the AI waste backlog.

Every significant repeated AI waste pattern must be evaluated as a candidate for workflow conversion.

Workflow Conversion, AI Retrospective, AI Capacity Planning.

26. Workflow Conversion

Definition

Workflow Conversion is the process by which a community converts repeated AI work into a non-AI workflow, validator, template, script, UI action, rule, or other deterministic capability.

Why it matters

AI should help discover repeatable patterns. When a pattern repeats, the community should consider converting it into a stable workflow.

This moves AI toward higher-value work and makes routine work more reliable.

Minimum requirement

Repeated AI work must be reviewed during retrospective to determine whether it should be converted into a non-AI workflow.

AI Waste Backlog, Skill Evolution, Human Capability Reserve.

27. Human Skill

Definition

Human Skill is structured knowledge intended for people so they can perform work in line with the community’s values, quality, and style.

It may contain:

Why it matters

An AI-first community must not create know-how only for AI.

People must remain able to learn, work, and validate AI outputs.

Minimum requirement

Critical community capabilities must have a human-usable form, especially when they are also supported by AI.

AI Skill, Skill Evolution, Human Capability Reserve.

28. AI Skill

Definition

AI Skill is a structured instruction or set of rules by which an AI agent performs a specific type of work according to the community standard.

It may contain:

Why it matters

AI skills enable repeatable and governed AI behavior.

If AI skills are stored only in a proprietary tool, skill lock-in emerges.

Minimum requirement

Critical AI skills must be exportable, versioned, and stored in the Source of Truth or derivable from it.

Human Skill, Skill Evolution, AI Lock-in, Agent Permissions.

29. Skill Evolution

Definition

Skill Evolution is the process in which work experience is converted into updated human skills and AI skills.

Why it matters

AI should not only execute work. It should help extract know-how from work.

When a good output appears, the community should ask which rule or pattern can be learned from it.

Minimum requirement

Significant new know-how, lessons learned, and good outputs must be evaluated as candidates for skill updates.

A skill update must be approved by responsible human or community governance.

AI Retrospective, Human Skill, AI Skill, Feedback Loop.

30. Feedback Loop

Definition

Feedback Loop is the mechanism by which experience, signals, risks, opportunities, and change proposals flow from work back into decisions, strategy, values, workflows, and skills.

Why it matters

AIFC is not only a top-down system.

Values and purpose flow downward. Experience, signals, and change proposals flow upward.

Without a feedback loop, the community does not learn and cannot adapt its behavior to reality.

Minimum requirement

An AIFC community must have a mechanism for collecting, structuring, evaluating, and deciding on change proposals.

Change Proposal, AI Retrospective, Decision Record, Community Interface.

31. Change Proposal

Definition

Change Proposal is a structured proposal to change direction, strategy, value interpretation, workflow, skill, governance, interface, or another community element.

It may come from:

Why it matters

A change proposal enables bottom-up adaptation of the community.

AI may detect a signal and formulate a proposal, but the proposal is not a decision.

Minimum requirement

Significant change proposals must be:

Feedback Loop, Decision Record, Values, Community Interface.

32. Observed Signal

Definition

Observed Signal is an event, trend, repeated problem, opportunity, risk, or other reality input that may trigger a change proposal.

A signal may come from:

Why it matters

The community must be able to respond to reality, not only execute a predefined plan.

AI may be useful in detecting signals people would miss.

Minimum requirement

It must be possible to record significant signals and convert them into a change proposal or Decision Record.

Feedback Loop, Change Proposal, AI Retrospective.

33. Decision Record

Definition

Decision Record is a record of a community decision.

It contains:

Why it matters

Without Decision Records, the community loses the memory of its own learning.

Decision Records connect feedback, change, values, and responsibility.

Minimum requirement

Critical decisions and accepted significant change proposals must have a Decision Record.

Source of Truth, Feedback Loop, Change Proposal, Auditability.

34. Community Interface

Definition

Community Interface is the standardized way a community describes itself, its boundaries, inputs, outputs, needs, offers, values, decision rules, and cooperation with other communities.

Why it matters

Communities do not exist in isolation.

An interface enables cooperation between teams, departments, companies, states, and wider systems.

Minimum requirement

An AIFC community must be able to describe:

Community, Shared Values Layer, Multi-Community Governance, Feedback Loop.

35. Shared Values Layer

Definition

Shared Values Layer is a set of values or principles shared by multiple communities that enables their coordination, cooperation, and conflict resolution.

Why it matters

When a community scales to a higher level, such as a company as a community of departments or a state as a community of communities, it needs a shared values layer.

Minimum requirement

Multiple connected AIFC communities must have a way to define, share, and evaluate common values.

Values, Community Interface, Multi-Community Governance.

36. Company as a System

Definition

Company as a System is the application of AIFC principles to a company.

A company is described as an operable system containing:

Why it matters

A company is not only a legal entity, people, and software.

It can be described as a system that can be understood, audited, improved, replicated, licensed, or launched.

Minimum requirement

An AIFC-compatible company must have a defined operating model, Source of Truth, responsibility, values, AI governance, and fallback for critical capabilities.

Operational DNA, Company as Product, Company Generation, Ghost AI Company.

37. Company as Product

Definition

Company as Product is the concept that a well-described company operating model may be sold, licensed, forked, localized, or used as the basis for launching a new company.

Why it matters

Structured know-how turns a company into a replicable organizational system.

This increases company value, but also increases security risk if Operational DNA leaks.

Minimum requirement

If a community treats a company as a product, it must protect its Operational DNA and define responsibility, values, license, security, and limits of replication.

Company as a System, Operational DNA, Knowledge Security, Ghost AI Company.

38. Company Generation

Definition

Company Generation is the process in which a new company is first designed as an AIFC-compatible operating system and only then launched as a legal, business, and human organization.

It may include:

Why it matters

AI makes it possible to design companies faster and more systematically than before.

Without AIFC, this may create ghost AI companies or AI-dependent operating models.

Minimum requirement

A generated company must have identifiable human or community ownership of purpose, values, and responsibility.

Company as a System, Company as Product, Ghost AI Company.

39. Digital Company

Definition

Digital Company is a company whose primary operating model, services, workflows, communication, and work are mainly digital and can be performed to a high degree by people and AI agents.

Why it matters

A digital company may be optimized for digital employees, AI agents, and workflow APIs.

This can significantly reduce operating costs, but requires strong governance and transparent responsibility.

Minimum requirement

A digital company must have a clear human or community owner, audit, responsibility, fallback, values, and rules for AI autonomy.

Company as a System, AI Intensity, Ghost AI Company.

40. Ghost AI Company

Definition

Ghost AI Company is an organization that has the external facade of a company, but lacks a real responsible community, values, fallback, ownership of purpose, and human or community governance.

It may have:

But it lacks responsibility.

Why it matters

AI has reduced the cost of creating the appearance of a company almost to zero.

This creates the risk of organizations that look like companies but are not responsible communities inside.

Minimum requirement

AIFC rejects the ghost AI company model.

Every AI-first company must have identifiable human or community ownership of purpose, values, critical decisions, and responsibility.

Human Ownership, Company Generation, Digital Company, Company as a System.

41. Agent

Definition

Agent is an AI or software entity able to perform tasks according to instructions, access to tools, data, workflows, and rules.

An agent may:

Why it matters

An AI agent with tools and access to data is not only a chatbot. It is an active work entity.

It therefore needs identity, permissions, scope, and audit.

Minimum requirement

Every agent working with non-public or critical know-how must have defined:

Agent Permissions, AI Skill, AI as Team Member, AI-NDA Boundary.

42. AI as Team Member

Definition

AI as Team Member is the concept in which an AI agent acts as a managed team member with a role, tasks, permissions, limits, review, and a responsible human owner.

Why it matters

This concept helps companies understand AI as work capacity in a team, not only as a tool.

At the same time, it must not hide that AI is not a bearer of human responsibility.

Minimum requirement

An AI team member must have:

Agent, Human Ownership, AI Capacity, AI Retrospective.

43. Agent Permissions

Definition

Agent Permissions are rules that determine what an agent may read, change, run, propose, or decide.

Why it matters

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

Like people, agents must follow least privilege, need to know, auditability, and revocation.

Minimum requirement

An agent with access to non-public data or tools must have defined and auditable permissions.

AI-NDA Boundary, Agent, Access Control, Auditability.

44. Auditability

Definition

Auditability is the ability to trace who or what performed an action, based on which inputs, with which permission, with which output, and who approved the result.

Why it matters

Without auditability, responsibility, security, quality, and learning cannot be managed.

AI agentic actions must be auditable, especially when they affect the knowledge base, workflows, customers, finance, security, or decisions.

Minimum requirement

Critical AI workflows must have an audit trail.

Decision Record, Agent Permissions, Knowledge Security, AI-NDA Boundary.

45. Data Classification

Definition

Data Classification is the division of data and know-how by sensitivity and usage rules.

AIFC recommends these minimal layers:

Why it matters

Without classification, the community cannot determine what AI may see, process, or store.

Minimum requirement

The community must classify know-how before making it available to AI tools or agents.

AI-NDA Boundary, Operational DNA, Knowledge Security.

46. Knowledge Security

Definition

Knowledge Security is the protection of community knowledge, skills, workflows, decisions, metadata, and Operational DNA.

Why it matters

In an AI-first community, the knowledge base may be the most valuable asset.

A knowledge base leak may mean leaking the community’s ability to operate, compete, or replicate itself.

Minimum requirement

The knowledge base must have:

Operational DNA, Source of Truth, AI Lock-in.

47. AIFC Compliance

Definition

AIFC Compliance is the degree to which a community satisfies AIFC principles and requirements.

Compliance may have levels, for example:

Why it matters

Communities need to measure where they are and which next step makes sense.

Minimum requirement

Minimal AIFC compliance requires:

Minimal AIFC Compliance, AIFC Manifest, Governance.

48. Minimal AIFC Compliance

Definition

Minimal AIFC Compliance is the basic set of requirements without which a community cannot be considered a human-managed AI-first community.

Minimum requirement

The community must at least:

  1. Define its purpose.
  2. Define its values.
  3. Have a human or community owner of responsibility.
  4. Have a knowledge base as the Source of Truth.
  5. Distinguish public, internal, restricted, and critical know-how.
  6. Define an AI-NDA Boundary for non-public data.
  7. Plan AI as limited capacity.
  8. Manage AI autonomy according to risk, budget, and governance.
  9. Have fallback for critical AI workflows.
  10. Ensure that know-how created with AI returns to the Source of Truth.
  11. Regularly evaluate AI benefit, cost, noise, and dependency.
  12. Convert repeated AI routine work into non-AI workflows where it makes sense.
  13. Develop human skills and AI skills.
  14. Maintain human ability to perform work without AI.
  15. Protect Operational DNA as a critical asset.
  16. Have an exit strategy for AI vendors, models, agents, and proprietary skill stores.
  17. Reject the Ghost AI Company model without a responsible community.
  18. Allow community members to propose changes to direction, strategy, workflows, skills, or governance.
  19. Allow authorized AI agents to propose changes based on detected signals, risks, or opportunities.
  20. Ensure that a change proposal is not automatically a decision.
  21. Record accepted and rejected changes in Decision Records or the Source of Truth.

AIFC Compliance, AIFC Manifest, Core Concepts.

49. Summary

AIFC defines language for human-managed AI-first communities.

The core concepts create a shared model:

Community
-> Purpose
-> Values
-> Knowledge Base
-> Operational DNA
-> Human-managed AI
-> AI Capacity
-> AI-NDA Boundary
-> Feedback Loop
-> Skill Evolution
-> Community Interface
-> Company as a System

AIFC rests on a simple distinction:

AI may accelerate the community. AI must not own its purpose.

AI may create outputs. Know-how must return to the Source of Truth.

AI may propose changes. The community decides.

AI may strengthen a company. It must not turn it into a Ghost AI Company.

AI-first. Human-managed. Purpose-driven. Feedback-enabled.