AIFC-042: Workflow Conversion

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

AIFC-000 Manifest of the AI-first community
AIFC-001 Core Concepts
AIFC-002 Community Model
AIFC-004 Feedback and Change Proposals
AIFC-010 Knowledge Structure
AIFC-012 Metadata and Markdown
AIFC-013 Human and AI Readable Content
AIFC-020 Human-Managed AI
AIFC-024 Human Capability Reserve
AIFC-030 AI Capacity Planning
AIFC-033 AI Budget and Cost Control
AIFC-040 AI Retrospective
AIFC-041 AI Waste Backlog

Purpose of this document: Define Workflow Conversion as the mechanism by which a community converts repeated AI work, AI waste, unclear routines, or manual repetition into stable, governed, and maintainable system capabilities: workflows, templates, validators, scripts, Human Cockpit actions, human skills, AI skills, or decisions to stop doing the work.

1. Purpose of this document

This document defines Workflow Conversion.

An AI-first community must not use AI forever for the same repeated routine when that routine can be converted into a more stable system solution.

AI can help discover a pattern.

But when the pattern repeats, the community should consider whether it should become:

workflow,
template,
validator,
script,
automation,
Human Cockpit action,
metadata rule,
checklist,
human skill,
AI skill,
decision rule,
or a decision to stop doing the work.

Workflow Conversion is the mechanism by which repeated AI consumption becomes system capability.

2. Core principle

The core principle of this document is:

Repeated AI work should become reusable system capability.

AIFC says:

Use AI to discover patterns.
Do not pay AI forever to repeat patterns that the system can own.

AI is excellent for discovery, formulation, and transformation.

A stable system should capture repeated patterns, make them cheaper, make them clearer, and make them resilient.

3. Definition

Workflow Conversion is a managed process by which a community converts repeated or low-value AI work into a stable form of work or knowledge.

Workflow Conversion may convert an AI routine into:

standard workflow,
template,
validator,
deterministic script,
automation,
Human Cockpit UI action,
checklist,
metadata schema,
decision rule,
human skill,
AI skill,
training material,
or stop-work decision.

Workflow Conversion does not always mean more automation.

Sometimes it means less work, better structure, or a conscious decision that a given output is not needed.

Minimum requirement

A repeated AI routine with significant cost, review load, attention impact, or dependency risk must be assessed for Workflow Conversion.

4. Why Workflow Conversion matters

Without Workflow Conversion, AI may become an expensive and invisible glue between poorly defined parts of the system.

AI then repeatedly:

rewrites unclear inputs,
fixes inconsistent structure,
generates missing templates,
searches for nonexistent rules,
explains undocumented workflows,
creates outputs without owners,
bridges poor interfaces between people and the knowledge base.

This helps in the short term.

In the long term, it can hide that the system is not well designed.

Workflow Conversion prevents AI from becoming a permanent patch over a repeated structural problem.

Minimum requirement

An AI-first community must have a mechanism for extracting system improvement from repeated AI work.

5. Workflow Conversion vs automation

Workflow Conversion is not the same as automation.

Automation is one possible solution.

Workflow Conversion is broader.

It may lead to:

a better human procedure,
a clearer template,
a better knowledge base,
a smaller work scope,
a validator,
governance change,
a change in the AI role,
removal of an unnecessary output,
or simply a better decision rule.

Example

AI repeatedly summarizes long meeting notes.

Possible solutions:

meeting notes template,
clear Decisions, Actions, and Risks sections,
Human Cockpit action for extracting action items,
validator for missing owners,
AI only for unclear parts,
or a decision that some meetings do not need summaries.

Minimum requirement

Workflow Conversion must assess more solutions than simply “more automation.”

6. Sources of conversion candidates

Workflow Conversion candidates may come from:

AI Waste Backlog,
AI Retrospective,
cost review,
human review feedback,
repeated prompts,
repeated manual work,
agent logs,
support patterns,
knowledge cleanup,
validation failures,
Human Cockpit Layer signals,
maintenance backlog,
capability incidents,
lock-in assessment,
AI dependency signals.

Minimum requirement

Workflow Conversion must be connected at least to the AI Waste Backlog and AI Retrospective.

7. Conversion candidate

A Workflow Conversion candidate is an item that may be converted into a more stable system solution.

The candidate should describe:

repeated pattern,
current way of working,
AI role,
frequency,
cost,
review load,
attention impact,
dependency risk,
Source of Truth impact,
possible solution,
owner,
priority,
expected benefit.

Minimum requirement

A Workflow Conversion candidate must have an owner or triage owner and a clearly described pattern.

8. Conversion lifecycle

AIFC recommends this lifecycle:

observed
-> candidate
-> triaged
-> solution designed
-> approved
-> implemented
-> validated
-> documented
-> closed

Observed

The pattern was recorded.

Candidate

The pattern was recognized as a conversion candidate.

Triaged

Impact, cost, and priority were assessed.

Solution designed

A solution was designed.

Approved

The solution was approved by the responsible role.

Implemented

The solution was created.

Validated

The community verified that it reduced waste, risk, or load.

Documented

The solution was written into the Source of Truth.

Closed

The item was closed.

Minimum requirement

Significant Workflow Conversion must have a lifecycle status and a verification step.

9. Triage criteria

Triage assesses whether conversion is worth doing.

Criteria:

repetition frequency,
AI cost,
human review cost,
attention cost,
dependency risk,
security risk,
Source of Truth gap,
impact on values,
impact on customers,
impact on Operational DNA,
simplicity of the solution,
verifiability,
reuse potential,
potential to reduce debt.

Prioritization example

High priority:
Repeated pattern with high AI cost, high review load or dependency risk.

Medium priority:
Repeated pattern with moderate cost and clear conversion path.

Low priority:
Rare pattern or unclear benefit. Monitor only.

Minimum requirement

Workflow Conversion candidates must be prioritized by impact and return.

10. Conversion solution types

AIFC distinguishes several solution types.

10.1 Template conversion

A repeated output is converted into a template.

Example:

Jira ticket template,
decision record template,
meeting notes template,
change proposal template.

10.2 Validator conversion

A repeated check is converted into a validation rule.

Example:

every critical workflow must have an owner,
every AI workflow must have a fallback,
restricted content must not have unrestricted AI access.

10.3 Script conversion

A routine transformation is converted into a script.

Example:

file renaming,
metadata key check,
export of ownerless artefacts.

10.4 Workflow conversion

Unclear repeated work is converted into a formal workflow.

Example:

review AI-generated change proposal,
approval process,
Source of Truth write-back.

10.5 Human Cockpit action

Repeated work is converted into a UI action or guided dialogue.

Example:

“Create change proposal from observed signal”
“Approve AI-generated skill update”
“Move block to correct metadata folder”

10.6 Human skill conversion

Experience is converted into a human skill.

Example:

how to write a good Jira ticket,
how to review an AI output,
how to run a fallback manually.

10.7 AI skill conversion

A repeated good pattern is converted into an AI skill.

Example:

UX review agent skill,
backlog refinement agent skill,
documentation cleanup agent skill.

10.8 Decision rule conversion

Repeated decision-making is converted into a rule.

Example:

when to use Conservative Mode,
when to require the AI-NDA Boundary,
when an AI output must be reviewed.

10.9 Source of Truth conversion

Missing knowledge is written as a knowledge artefact.

Example:

FAQ,
workflow document,
concept definition,
glossary,
Decision Record.

10.10 Stop-work conversion

The community decides that the work will not be done.

Example:

regular AI report that nobody uses,
summary without an audience,
proposal generation without an owner.

Minimum requirement

Workflow Conversion must explicitly choose a solution type.

11. When not to convert

Not every AI-assisted task should be converted.

AI may remain appropriate when the work:

is creative,
is one-off,
is highly contextual,
requires synthesis from new sources,
has low frequency,
has high variability,
has low cost,
has no clear pattern,
or has conversion cost greater than expected benefit.

Example

AI helps formulate strategic options for a unique decision.

That may not be appropriate for automation.

But it may still produce a human skill or decision support template.

Minimum requirement

Workflow Conversion must support the decision do not convert or monitor only.

12. Conversion and Source of Truth

Every significant conversion must end in the Source of Truth.

If it creates:

template,
workflow,
validator,
skill,
rule,
checklist,
Human Cockpit action,
decision rule,

then it must be discoverable who owns it, what status it has, and where it is used.

Otherwise, conversion creates another hidden mechanism.

Minimum requirement

Significant conversion outputs must be written into the Source of Truth or an appropriate versioned repository.

13. Conversion and Human Capability Reserve

Workflow Conversion may strengthen or weaken human capability.

It strengthens it when it:

creates a human skill,
improves onboarding,
preserves explanation,
improves review,
reduces routine load while preserving understanding,
creates a fallback.

It weakens it when it:

hides decision logic,
removes learning tasks,
leaves people only clicking,
moves capability into invisible automation,
lacks human-readable documentation.

Minimum requirement

Workflow Conversion of critical work must assess impact on Human Capability Reserve.

14. Conversion and AI dependency

One goal of conversion is to reduce AI dependency.

Example:

Before:
Team needs AI for every Jira ticket rewrite.

After:
Standard ticket template + examples + AI only for complex tickets.

Dependency decreases because the system now holds the pattern.

AI remains support, not a mandatory intermediary.

Minimum requirement

Workflow Conversion must consider whether the result reduces or increases AI dependency.

15. Conversion and AI skills

Workflow Conversion may lead to the creation or update of an AI skill.

An AI skill is appropriate when:

the work remains variable,
AI is still the most suitable executor,
but it needs better instructions,
better role,
better output format,
better forbidden actions,
or better examples.

An AI skill must not become the only carrier of critical know-how.

Minimum requirement

A critical AI skill created through Workflow Conversion must be connected to human-readable knowledge or a human skill.

16. Conversion and human skills

Workflow Conversion may lead to a human skill.

A human skill is appropriate when:

people repeatedly assign work to AI poorly,
people cannot review outputs,
people cannot create the basic output without AI,
people cannot use the template,
people cannot decide when to use AI.

Minimum requirement

If conversion addresses AI dependency or human review weakness, it must assess the need for a human skill update.

17. Conversion and validation rules

If AI repeatedly checks the same condition, a validator may be appropriate.

Examples:

missing owner,
missing status,
missing review date,
AI workflow without fallback,
restricted artefact with ai_access: allowed,
Decision Record without related artefact,
AI team member without owner.

A validator is often better than AI because it is:

cheaper,
deterministic,
fast,
auditable,
consistent,
less prone to hallucination.

Minimum requirement

A repeated AI check of a structural rule must be assessed for validator conversion.

18. Conversion and Human Cockpit Layer

The Human Cockpit Layer is often the best place to convert repeated work into a human-usable action.

Examples:

“Approve / reject AI-generated proposal”
“Create decision record from accepted proposal”
“Mark as dependency risk”
“Move block to correct classification”
“Generate fallback checklist”
“Create human skill from repeated review correction”

A Human Cockpit action may combine:

AI assistance,
metadata,
human approval,
validation,
write-back into the Source of Truth.

Minimum requirement

Workflow Conversion must assess whether repeated work belongs in the Human Cockpit Layer.

19. Conversion and cost control

Workflow Conversion is one of the main tools of AI cost control.

It reduces costs because:

AI is not used for routine work,
simple checks are handled by validators,
templates reduce prompting,
a better Source of Truth reduces repeated explanation,
Human Cockpit actions reduce manual coordination,
better AI skills reduce rejection rate.

Minimum requirement

High-cost AI waste items must be assessed for Workflow Conversion.

20. Conversion and quality

Workflow Conversion should improve quality, not only reduce costs.

It can improve quality by:

standardizing inputs,
standardizing outputs,
adding quality gates,
adding examples,
reducing variability,
improving review,
preventing repeated errors,
improving the Source of Truth.

Minimum requirement

Workflow Conversion must define how the quality of the resulting solution will be verified.

21. Conversion and governance

Workflow Conversion may change governance.

For example, if AI repeatedly proposes the same low-risk changes, the community may introduce:

delegated approval,
batch approval,
automated low-risk handling,
explicit threshold,
or lower review depth.

Conversely, if AI creates risk, conversion may make governance stricter.

Minimum requirement

Workflow Conversion must assess whether the proposed solution changes approval rules, AI Autonomy, or operating mode.

22. Conversion and maintenance

Workflow Conversion may create a new artefact that needs care.

Every new template, validator, workflow, or skill must have:

owner,
status,
review cycle,
maintenance rules,
Source of Truth location,
versioning,
retirement mechanism.

Otherwise, conversion only moves debt to another place.

Minimum requirement

Every significant conversion output must have an owner and review mechanism.

23. Conversion pattern examples

Example 1: Repeated Jira ticket rewriting

Observed pattern:

AI repeatedly rewrites unclear Jira tickets into a better structure.

Possible conversion:

standard Jira ticket template,
examples of good tickets,
human skill for ticket writing,
AI skill for unclear ticket improvement,
validator for missing acceptance criteria.

Example 2: Repeated metadata cleanup

Observed pattern:

AI repeatedly detects missing metadata in Markdown artefacts.

Possible conversion:

metadata schema,
static validator,
Cockpit warning,
cleanup agent only for ambiguous cases.

Example 3: Repeated meeting summarization

Observed pattern:

AI summarizes meetings into decisions and action items.

Possible conversion:

meeting notes template,
decision/action sections,
action owner metadata,
AI extraction only for unstructured notes.

Example 4: Repeated AI review corrections

Observed pattern:

Human reviewers repeatedly fix the same style and structure mistakes in AI outputs.

Possible conversion:

AI skill update,
output style rules,
examples,
forbidden patterns,
human review checklist.

Example 5: Repeated strategic explanation

Observed pattern:

AI repeatedly explains the same strategy to new team members.

Possible conversion:

strategy summary,
onboarding artefact,
Human Cockpit card,
FAQ,
Source of Truth update.

24. Conversion validation

After conversion is implemented, impact must be verified.

Possible measures include:

reduced AI cost,
reduced token usage,
reduced review time,
reduced rejection rate,
fewer repeated prompts,
higher input quality,
higher output quality,
reduced dependency,
better availability of know-how,
faster onboarding,
lower attention load.

Minimum requirement

Significant Workflow Conversion must have a predefined validation method.

25. Conversion record

Workflow Conversion should have a record.

The record should include:

original pattern,
reason for conversion,
selected solution type,
rejected alternatives,
owner,
expected benefit,
impact on AI usage,
impact on human capability,
impact on governance,
implementation,
validation,
Source of Truth update.

Minimum requirement

Significant Workflow Conversion must be traceable as a Decision Record, change proposal, or conversion record.

26. Relationship with AI Waste Backlog

The AI Waste Backlog detects waste.

Workflow Conversion resolves it.

Relationship:

AI Waste Backlog
-> identifies repeated or low-value AI work

Workflow Conversion
-> converts it into stable system capability

Minimum requirement

High-priority AI Waste Backlog items must be assessed for Workflow Conversion.

27. Relationship with Skill Evolution

Workflow Conversion often leads to Skill Evolution.

If conversion creates or changes a human skill or AI skill, it must connect to AIFC-043 Skill Evolution.

Examples:

new dashboard UX skill,
better AI documentation skill,
human review skill,
prompt-to-change-proposal skill,
support escalation skill.

Minimum requirement

Workflow Conversion outputs that affect skills must create a skill update proposal.

28. Relationship with AI Retrospective

AI Retrospective is where conversion candidates often emerge.

The retrospective should ask:

What did we repeatedly do using AI?
What had low value?
What should become a template?
What should become a validator?
What should become a skill?
What should become a workflow?
What should we stop doing?

Minimum requirement

AI Retrospective must be able to create Workflow Conversion candidates.

29. Relationship with Feedback and Change Proposals

Workflow Conversion is a system change.

It may therefore be processed as a change proposal.

Especially when it changes:

workflow,
governance,
roles,
Source of Truth,
AI Autonomy,
Human Cockpit Layer,
validation rules,
team responsibilities.

Minimum requirement

Workflow Conversion with significant impact must go through a change proposal or decision mechanism.

30. Relationship with Human Cockpit Layer

The Human Cockpit Layer may support the entire conversion lifecycle.

It may show:

conversion candidates,
AI waste origin,
proposed solution types,
expected savings,
dependency impact,
owner,
approval state,
implementation state,
validation result,
related Source of Truth update.

Minimum requirement

Responsible roles must have a human-accessible view of significant Workflow Conversion candidates and their status.

31. AI role in Workflow Conversion

AI may help with conversion.

It may:

find repeated patterns,
propose solution types,
prepare a template,
propose a validator,
draft a workflow,
prepare a human skill,
prepare an AI skill,
estimate cost savings,
propose a validation method.

AI must not decide by itself that a specific activity should be stopped, automated, or converted when that has governance, values, or organizational impact.

Minimum requirement

AI-generated conversion proposals must be marked as proposals and reviewed by an owner.

32. Suggested metadata

Example metadata for a Workflow Conversion candidate:

workflow_conversion_candidate:
  id:
  title:
  status: observed | candidate | triaged | solution_designed | approved | implemented | validated | documented | closed | rejected | monitor_only
  owner:
  triage_owner:
  source:
    - ai_waste_backlog
    - ai_retrospective
    - cost_review
    - human_review_feedback
    - cockpit_signal
    - incident_review
    - team_feedback
  observed_pattern:
  current_ai_usage:
  related_ai_waste_items:
  related_ai_workflows:
  related_ai_team_members:
  frequency:
  cost_impact: low | medium | high | critical
  review_impact: low | medium | high | critical
  attention_impact: low | medium | high | critical
  dependency_impact: low | medium | high | critical
  proposed_solution_type:
    - template
    - validator
    - script
    - workflow
    - human_cockpit_action
    - human_skill
    - ai_skill
    - decision_rule
    - source_of_truth_update
    - stop_work
    - monitor_only
  expected_benefit:
  human_capability_impact:
  governance_impact:
  selected_solution:
  rejected_alternatives:
  validation_method:
  related_change_proposal:
  related_decision_record:
  source_of_truth_update:
  created_at:
  last_reviewed:

This structure is illustrative.

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

33. Anti-patterns

AIFC rejects the following anti-patterns.

33.1 Permanent AI routine

AI performs a repeated pattern forever even though it should become a system capability.

33.2 Automation without understanding

A routine is automated, but nobody understands the rule that the automation performs.

33.3 Workflow conversion without owner

A template, script, or validator is created, but nobody owns it.

33.4 Conversion that hides decision logic

A decision rule is moved into a tool or agent without a human-readable description.

33.5 Conversion that increases dependency

The solution reduces cost but increases dependency on AI or a vendor.

33.6 Template nobody uses

A template is created, but it is not connected to a workflow or Human Cockpit Layer.

33.7 Validator without context

A rule is created, but people do not understand why it exists or when to use it.

33.8 Stop-work avoided

The community automates work that should not be done at all.

33.9 No validation

Conversion is implemented, but nobody verifies whether it reduced waste or increased value.

33.10 AI decides conversion alone

AI decides by itself that work should be automated, stopped, or converted without governance.

34. Minimal requirements

In the area of Workflow Conversion, an AIFC community must at minimum:

Have a mechanism for identifying conversion candidates.
Connect Workflow Conversion to the AI Waste Backlog.
Connect Workflow Conversion to AI Retrospective.
Ensure each conversion candidate describes a repeated pattern.
Ensure each conversion candidate has an owner or triage owner.
Prioritize conversion candidates by impact and return.
Assess more solution types than automation alone.
Support the decision do not convert or monitor only.
Write significant conversion outputs into the Source of Truth.
Assess Human Capability Reserve impact for conversion of critical work.
Assess impact on AI dependency.
Connect critical AI skills created through conversion to human-readable knowledge or a human skill.
Assess human skill updates when conversion addresses dependency or review weakness.
Assess repeated AI checks of structural rules for validator conversion.
Assess the possibility of a Human Cockpit action.
Assess high-cost AI waste items for Workflow Conversion.
Define how the quality of the resulting solution will be verified.
Ensure each conversion output has an owner and review mechanism.
Provide a validation method for significant conversion.
Make significant conversion traceable as a Decision Record, change proposal, or conversion record.
Create a skill update proposal when conversion affects skills.
Route high-impact conversion through a change proposal or decision mechanism.
Make significant conversion candidates visible through the Human Cockpit Layer or a governance interface.
Mark AI-generated conversion proposals as proposals and review them by an owner.

35. Summary

Workflow Conversion is the mechanism by which an AI-first community learns from repeated AI work.

AI is very good at helping reveal a pattern.

Once the pattern repeats, the community should consider whether the system should own it.

AIFC therefore says:

Let AI reveal the pattern.
Let the system own the pattern.

Workflow Conversion reduces AI waste, strengthens the Source of Truth, improves human skills, makes routine work cheaper, and reduces dependency.

The AI-first community does not lose AI through this process.

It gains the ability to use AI where it has higher value.

Workflow Conversion turns repeated AI work into durable community capability.