AIFC-030: AI Capacity Planning

Status: Draft 0.1 Standard: AI-First Community Standard Abbreviation: AIFC Builds on:

• AIFC-000 Manifesto of an AI-first community
• AIFC-001 Core Concepts
• AIFC-003 Values and Purpose
• AIFC-011 Operational DNA
• AIFC-020 Human-Managed AI
• AIFC-021 AI as External Expert Capacity
• AIFC-022 AI-NDA Boundary
• AIFC-024 Human Capability Reserve

Purpose of this document: Define AI Capacity Planning as the way an AIFC community plans, allocates, measures, and evaluates the limited capacity it has for AI work.

1. Purpose of this document

This document defines AI Capacity Planning.

AI capacity is not unlimited.

It is constrained by money, tokens, compute, models, human review time, governance attention, security capacity, risk tolerance, and the community’s ability to absorb AI-generated work.

AIFC therefore treats AI capacity as a governed community resource.

The goal is not to use as much AI as possible.

The goal is to use AI capacity where it strengthens purpose, values, learning, delivery, resilience, and the source of truth without creating unmanaged dependency.

2. Core principle

The core principle of this document is:

AI capacity must be planned as a scarce governed resource, not treated as unlimited acceleration.

AI can accelerate the community.

But acceleration consumes capacity and creates review, cost, security, and dependency obligations.

AIFC therefore says:

Plan AI capacity where it creates durable community value.
Measure its cost.
Protect human review capacity.
Convert useful outputs back into the source of truth.

3. Definition

AI Capacity Planning is the process of deciding how much AI capacity the community can safely and usefully use, where it should be allocated, how it should be measured, and when it should be reduced, redirected, or stopped.

AI capacity includes:

• financial budget,
• token and compute availability,
• model and tool availability,
• human review capacity,
• governance capacity,
• security capacity,
• attention capacity,
• risk capacity,
• Human Capability Reserve.

Minimum requirement

Significant AI use must be planned against available AI Capacity and available human review capacity.

4. Why AI Capacity Planning matters

Without capacity planning, AI use tends to expand where it is easiest, not where it is most valuable.

This creates several risks:

• AI budget is consumed by low-value work,
• human review becomes a bottleneck,
• AI-generated material accumulates without source-of-truth write-back,
• security and AI-NDA review are bypassed,
• AI dependency grows unnoticed,
• AI appears productive while creating maintenance debt.

Minimum requirement

The community must be able to explain where significant AI capacity is being used and why.

5. AI capacity is multi-dimensional

AI capacity has several dimensions.

5.1 Financial capacity

Money available for models, tools, agents, infrastructure, and vendors.

5.2 Token and compute capacity

The practical amount of model execution available to the community.

5.3 Human review capacity

The time and competence available to review AI outputs.

5.4 Governance capacity

The ability to approve, audit, prioritize, and adjust AI use.

5.5 Security capacity

The ability to assess AI access, AI-NDA Boundaries, incidents, and data risk.

5.6 Attention capacity

The human attention available to absorb, decide, and act on AI outputs.

5.7 Risk capacity

The amount of AI-related risk the community is willing and able to carry.

5.8 Human capability capacity

The ability of people to understand, validate, and continue work with reduced AI or without AI.

Minimum requirement

AI capacity planning must consider more than financial cost.

6. AI capacity and purpose

AI capacity should be allocated toward the community’s purpose.

High AI activity that does not support purpose is waste.

AI can help with exploration and experimentation, but sustained AI capacity should be tied to a purpose, outcome, risk reduction, learning goal, or source-of-truth improvement.

Minimum requirement

Significant AI capacity must have a stated purpose.

7. AI capacity and values

Values constrain AI capacity decisions.

The community must not allocate AI capacity in a way that violates the values it claims to protect.

For example, capacity planning must consider whether AI use:

• reduces human dignity or agency,
• weakens accountability,
• hides decisions,
• increases unfairness,
• creates avoidable dependency,
• exposes sensitive know-how,
• shifts risk to another community.

Minimum requirement

AI capacity allocation must be reviewable against community values.

8. Capacity allocation domains

AI capacity may be allocated across several domains.

8.1 Delivery / change

AI helps create, improve, migrate, or transform systems and artefacts.

8.2 Support

AI helps triage, summarize, answer, and detect support patterns.

8.3 Maintenance

AI helps detect drift, duplication, missing owners, broken links, and outdated knowledge.

8.4 Knowledge management

AI helps structure, summarize, validate, and return know-how to the source of truth.

8.5 Security and compliance

AI helps review risks, access, controls, incidents, and compliance evidence.

8.6 Learning and skill evolution

AI helps convert experience into human skills, AI skills, training material, and improved workflows.

Minimum requirement

The community should know the main domains where AI capacity is allocated.

9. AI capacity units

AIFC communities may define capacity units to make AI work easier to plan.

Examples:

• monthly AI budget,
• token pool,
• agent run count,
• review hours,
• approved AI engagement slots,
• high-risk AI access slots,
• model usage tiers,
• AI-assisted workflow capacity,
• maintenance scan capacity.

Example

ai_capacity_unit:
  type: review_hours
  period: monthly
  amount: 40
  owner: knowledge-governance
  allowed_use:
    - source_of_truth_review
    - ai_generated_change_proposals

Minimum requirement

If AI use becomes significant, the community must define a practical unit for tracking it.

10. AI capacity in sprint planning

For teams that use sprint or iteration planning, AI capacity should be planned alongside human capacity.

AI may reduce effort in one area while increasing:

• review work,
• prompt preparation,
• verification,
• source-of-truth updates,
• governance decisions,
• incident risk,
• cleanup.

Minimum requirement

AI-assisted work must not be planned as free capacity.

11. Human review capacity

Human review capacity is often the limiting factor.

AI may generate more proposals, drafts, reports, and warnings than humans can responsibly review.

If review capacity is missing, the answer is not automatically higher AI autonomy.

The answer may be:

• reduce AI intensity,
• prioritize outputs,
• narrow scope,
• improve review workflows,
• increase trained review capacity,
• defer low-value AI work.

Minimum requirement

AI capacity planning must include human review capacity.

12. Attention capacity

Attention is a scarce capacity.

AI can create too many signals, options, drafts, summaries, and recommendations.

If the community cannot attend to AI output, AI becomes noise rather than acceleration.

Minimum requirement

AI output volume must be limited to what the community can meaningfully process.

13. AI budget

AI budget is the financial and operational boundary for AI use.

It may include:

• money,
• tokens,
• model calls,
• infrastructure,
• vendor subscriptions,
• review time,
• governance time,
• security work.

Minimum requirement

Significant AI use must have budget visibility.

14. Budget thresholds

AI budgets should have threshold rules.

Examples:

80 % budget used -> notify owner
90 % budget used -> reduce non-critical AI work
100 % budget used -> switch to restricted mode

Thresholds protect the community from sudden outage and uncontrolled cost.

Minimum requirement

Critical AI capacity must define what happens when budget thresholds are reached.

15. AI capacity and operating modes

AI operating modes shape capacity use.

Conservative mode

Lower AI Intensity, lower autonomy, stronger review, lower risk.

Balanced mode

Normal AI use with review gates for significant actions.

Aggressive mode

Higher AI involvement for approved scope and time, with stronger monitoring.

Mission mode

Temporary high AI capacity for a specific goal, budget, owner, and exit.

Emergency AI-off mode

AI capacity is reduced or stopped to protect the community during incident, risk, budget, or trust failure.

Minimum requirement

AI capacity planning must be compatible with the community’s current operating mode.

16. AI capacity and autonomy

Higher AI autonomy usually requires more governance capacity, auditability, and fallback.

Autonomy should not be increased merely because human capacity is overloaded.

Example

Low review capacity
-> reduce AI output or prioritize review
-> do not silently raise autonomy for risky actions

Minimum requirement

AI autonomy must be adjusted according to risk and review capacity, not only according to speed pressure.

17. AI capacity and risk budget

Some AI use consumes risk capacity.

Examples:

• access to restricted data,
• work over Operational DNA,
• customer-facing output,
• agentic execution,
• high autonomy,
• external vendor dependency.

Minimum requirement

High-risk AI use must consume an explicit or implicit risk budget.

18. AI capacity and AI-NDA Boundary

AI capacity over non-public know-how must respect the AI-NDA Boundary.

The community should not allocate AI capacity to work that lacks approved data access, purpose limitation, storage rules, and owner.

Minimum requirement

AI capacity over non-public data must reference an approved AI-NDA Boundary.

19. AI capacity and Operational DNA

Operational DNA is a high-value area.

AI capacity used on Operational DNA may create high value, but also high risk.

It requires:

• explicit purpose,
• approved owner,
• limited scope,
• audit,
• AI-NDA Boundary,
• source-of-truth write-back,
• exit strategy.

Minimum requirement

AI capacity used on Operational DNA must be explicitly governed.

20. AI capacity and Human Capability Reserve

AI capacity must not silently replace human capability.

If all capacity is allocated to AI-assisted execution and none to human learning, fallback, review, or skill maintenance, the community becomes fragile.

Minimum requirement

Capacity planning must include protection of Human Capability Reserve.

21. AI waste

AI waste is AI capacity consumed without meaningful value.

Examples:

• repeated prompts that produce no accepted change,
• large summaries nobody reads,
• AI-generated documents not returned to the source of truth,
• automated reports without decisions,
• low-value model calls,
• agent work that creates cleanup burden.

Minimum requirement

The community must detect and reduce AI waste.

22. Workflow conversion

When repeated AI work proves useful, it should be converted into a better workflow.

This may mean:

• human skill,
• AI skill,
• checklist,
• template,
• automation,
• source-of-truth update,
• reusable review process.

Minimum requirement

Repeated valuable AI work should be assessed for workflow conversion.

23. Capacity planning cycle

AIFC recommends a capacity planning cycle:

plan
allocate
execute
measure
retrospect
convert / adjust

Plan

Define purpose, capacity, risks, owners, and expected value.

Allocate

Assign capacity to domains, workflows, teams, or engagements.

Execute

Use AI inside approved boundaries.

Measure

Measure cost, value, review load, risk, and source-of-truth return.

Retrospect

Review waste, dependency, risk, and learning.

Convert / adjust

Convert useful patterns into workflows or reduce low-value use.

Minimum requirement

Significant AI capacity must be reviewed periodically.

24. Capacity planning roles

AI capacity planning may involve:

• community owner,
• AI capacity owner,
• budget owner,
• security owner,
• knowledge owner,
• workflow owner,
• human review owner,
• Human Cockpit Layer owner.

Minimum requirement

Significant AI capacity must have an accountable owner.

25. Human Cockpit Layer and AI Capacity

The Human Cockpit Layer should make AI capacity visible.

It may show:

• active AI engagements,
• budget usage,
• review capacity,
• AI waste,
• capacity allocation,
• risk level,
• AI-NDA Boundaries,
• pending approvals,
• source-of-truth write-back status,
• dependency indicators.

Minimum requirement

The community must have a human-accessible way to see significant AI capacity use.

26. AI capacity record

AIFC recommends using an AI capacity record.

Example metadata:

ai_capacity:
  id:
  title:
  status: draft | active | paused | exhausted | under_review | archived
  owner:
  period:
  purpose:
  allocation_domain:
  budget_limit:
  token_limit:
  review_capacity:
  governance_capacity:
  risk_level:
  operating_mode:
  ai_nda_boundary:
  source_of_truth_targets:
  value_metrics:
  waste_indicators:
  review_cycle:
  last_reviewed:

This structure is illustrative.

The final schema should be defined in the agent-actionable layer of the standard.

27. Metrics

AI capacity should be measured across several metric types.

27.1 Cost metrics

• money spent,
• tokens used,
• model calls,
• compute,
• review hours,
• governance time.

27.2 Value metrics

• accepted proposals,
• reduced debt,
• improved workflows,
• knowledge returned to the source of truth,
• time saved,
• quality improved.

27.3 Risk metrics

• incidents,
• AI-NDA exceptions,
• high-risk engagements,
• AI dependency indicators,
• lock-in indicators.

27.4 Learning metrics

• skills updated,
• workflows converted,
• retrospectives completed,
• human capability improved.

Minimum requirement

AI capacity metrics must include value and risk, not only usage.

28. Anti-patterns

AIFC rejects the following anti-patterns.

28.1 Unlimited AI assumption

The community treats AI as if it has no cost, no limits, and no review burden.

28.2 AI budget without purpose

AI capacity is allocated without a clear reason.

28.3 AI capacity without review capacity

AI output grows faster than the community can review it.

28.4 AI usage as productivity theater

The community measures activity instead of durable value.

28.5 AI waste ignored

Low-value AI use continues because it looks modern or busy.

28.6 AI capacity used only for delivery

All capacity goes to output while maintenance, learning, and source-of-truth care are neglected.

28.7 Budget outage stops critical work

Critical workflows stop when budget or tokens run out.

28.8 Autonomy increased because review capacity is missing

AI is allowed to act more independently because humans are too busy to review it.

28.9 Hidden AI cost

Costs are spread across tools, teams, and review work so nobody sees the full picture.

28.10 No retrospective

The community uses AI but does not learn from the use.

29. Minimal requirements

In the area of AI Capacity Planning, an AIFC community must at minimum:

1. Treat AI capacity as limited.
2. Define a purpose for significant AI capacity.
3. Consider financial, token, compute, review, governance, security, attention, risk, and human capability capacity.
4. Allocate AI capacity according to purpose and values.
5. Track significant AI capacity in a practical unit.
6. Include AI capacity in planning when AI materially affects work.
7. Include human review capacity.
8. Define budget visibility.
9. Define budget thresholds for critical AI use.
10. Align capacity use with operating modes.
11. Govern autonomy according to risk and review capacity.
12. Respect AI-NDA Boundaries.
13. Govern AI capacity used on Operational DNA explicitly.
14. Protect Human Capability Reserve.
15. Detect AI waste.
16. Convert repeated useful AI work into workflows, skills, or source-of-truth updates.
17. Review significant AI capacity periodically.
18. Make significant AI capacity visible to accountable humans.

30. Summary

AI Capacity Planning turns AI use from uncontrolled consumption into governed community resource management.

AI capacity is powerful, but it is not free.

It consumes money, tokens, review time, governance attention, security capacity, and human capability.

AIFC therefore says:

Plan AI capacity.
Use it for purpose.
Measure value, not only usage.
Protect review capacity.
Reduce waste.
Return knowledge to the source of truth.
Keep the community capable without AI.

AI Capacity Planning turns AI acceleration into accountable capacity.