MWMS AI Brain Audit And Decay Prevention Framework

System: MWMS
Document Type: Framework
Authority Level: MCR Source Of Truth
Status: Draft For MCR
Primary Location: MCR
Future Operational Destination: HeadOffice Brain, AI Business Systems Brain, AI Manager, AI Employee Router, Brain Room, Content Brain, Offer Brain, Sales Brain, Creative Brain, Course Absorption System, Future AIBS Client Systems
Parent Page: HeadOffice
Owner: Martyn
Developer Boundary: No Development Action Authorized By This Page
Source Of Truth: MCR

Purpose

The purpose of this document is to define the MWMS AI Brain Audit And Decay Prevention Framework.

This framework establishes how MWMS detects, prevents, reviews, and corrects decay inside AI Brains, context libraries, skills, AI Employee procedures, offer files, voice systems, customer language banks, and client-specific AI systems.

MWMS will not remain accurate simply because a Brain, file, skill, or context library was once created correctly.

AI systems decay when:

offers change

markets shift

buyer language changes

old claims become unsafe

skills overlap

voice files become stale

client context changes

old positioning remains active

new proof is not added

retired language keeps resurfacing

AI Employees keep repeating old instructions

context libraries become bloated or duplicated

manual corrections are not turned into system updates

This framework exists to ensure MWMS treats AI Brains and AI skills as living systems, not static documents.

The goal is to prevent useful AI infrastructure from slowly becoming outdated, noisy, unsafe, or unreliable.

Scope

This framework applies to all MWMS Brains, AI Employees, context libraries, skill systems, offer libraries, client libraries, and future AIBS client systems.

This includes:

HeadOffice Brain

AI Business Systems Brain

AI Manager

AI Employee Router

Brain Room

Course Absorption System

Newsletter Intelligence

Opportunity System

Offer Brain

Content Brain

Creative Brain

Sales Brain

Conversion Brain

Customer Brain

Research Brain

Affiliate Brain

Ads Brain

Product Brain

Strategy Brain

Operations Brain

future AIBS client systems

This framework applies to:

AI Brain audits

context library audits

skill audits

voice audits

offer context audits

client Brain audits

customer language audits

retired language audits

proof library audits

AI Employee procedure audits

workflow drift reviews

manual correction reviews

This framework does not authorize technical development, plugin changes, Supabase changes, WordPress changes, automation wiring, or M developer action.

Core Definition

AI Brain Decay refers to the gradual loss of accuracy, usefulness, alignment, safety, or operational value inside an AI system.

Decay can occur in:

context files

skills

AI Employee instructions

voice rules

offer profiles

proof libraries

customer language banks

folder structures

workflow procedures

Brain routing rules

client context

manual prompts

audit records

Decay is not always obvious.

A system may still produce polished outputs while becoming less accurate.

The warning sign is not whether the output looks good.

The warning sign is whether the output still reflects current source truth, current strategy, current offer logic, current buyer reality, and current MWMS governance.

Core Principle

The core principle of this framework is:

Every AI Brain, skill, and context library must be reviewed before drift becomes operational damage.

MWMS must not wait until a system completely fails.

Small decay should be detected early.

Small corrections should be made before they compound.

This aligns with the MWMS Kaizen philosophy.

AI Brain audit work is not cleanup for its own sake.

It is ecosystem protection.

Primary Decay Types

MWMS recognizes ten major types of AI Brain decay.

  1. Context Decay

Context Decay occurs when source files are no longer current.

Examples:

offer details changed

price changed

buyer profile changed

positioning changed

proof changed

methodology changed

old campaign notes remain active

customer language is outdated

Risk:

AI Employees create outputs from stale context.

  1. Skill Decay

Skill Decay occurs when a reusable AI skill no longer follows the best current procedure.

Examples:

skill uses outdated rules

skill misses new standards

skill triggers incorrectly

skill produces old format

skill ignores new validation rules

skill overlaps with another skill

Risk:

Repeated work becomes repeatedly wrong.

  1. Voice Decay

Voice Decay occurs when AI output no longer sounds like the approved voice.

Examples:

copy becomes too polished

tone becomes generic

old phrases return

AI uses words the founder dislikes

brand becomes inconsistent

outputs from different Brains sound unrelated

Risk:

Trust and brand coherence weaken.

  1. Offer Decay

Offer Decay occurs when the AI system’s understanding of the offer no longer matches reality.

Examples:

old offer promise remains active

new inclusions are missing

old bonuses are mentioned

wrong audience is targeted

risk reversal changed

delivery model changed

Risk:

Public-facing assets misrepresent the offer.

  1. Proof Decay

Proof Decay occurs when proof assets become incomplete, outdated, unsupported, or risky.

Examples:

old testimonials remain active

case studies are no longer accurate

claims lack evidence

new proof is not added

compliance-sensitive claims are not flagged

Risk:

Sales, ads, and content use weak or unsafe credibility signals.

  1. Customer Language Decay

Customer Language Decay occurs when buyer wording, objections, anxieties, and desired outcomes are no longer current.

Examples:

new objections appear

old customer phrases are overused

buyer sophistication changes

market language shifts

support tickets reveal new friction

Risk:

Copy becomes disconnected from buyer reality.

  1. Folder Decay

Folder Decay occurs when the context library structure becomes messy.

Examples:

duplicate files

unclear source truth

drafts mixed with approved files

project files left active

archive files used accidentally

client files mixed with internal files

Risk:

AI Employees read the wrong files.

  1. Routing Decay

Routing Decay occurs when work no longer flows to the correct Brain, Employee, queue, or review path.

Examples:

course insights routed to wrong Brain

newsletter signals over-prioritized

offer decisions skip Research Brain

financial decisions skip Finance Brain

content outputs skip validation

Risk:

The system makes wrong operational decisions.

  1. Governance Decay

Governance Decay occurs when standards are not followed consistently.

Examples:

MCR pages use wrong structure

human review is skipped

change impact declarations are missing

citations or external course language leak into page output

M’s active boundary is ignored

Risk:

MCR becomes inconsistent and harder to govern.

  1. Client System Decay

Client System Decay occurs in future AIBS systems when client-specific context, skills, reports, or workflows become outdated.

Examples:

client offer changes

client approval rules change

client voice changes

client campaign goals change

client tool permissions change

client context leaks into another system

Risk:

Client-facing AI becomes unsafe, inaccurate, or untrustworthy.

Audit Triggers

MWMS should audit AI Brains, skills, and context libraries when any of the following occur.

Scheduled Trigger

quarterly review

monthly light check

after major campaign cycle

after launch

after major client delivery

after major course absorption block

after major system update

Performance Trigger

AI output keeps needing correction

output requires rewriting instead of polishing

content sounds generic

ads lose message clarity

reports become less useful

client feedback identifies mismatch

human operator keeps repeating the same correction

Context Trigger

offer changes

price changes

buyer profile changes

voice changes

new objections appear

new proof appears

compliance risk changes

old language is retired

new customer research is collected

System Trigger

new standard created

Brain rule changed

skill updated

workflow changed

tool permission changed

folder structure changed

AI Employee role changed

MCR page structure changed

Risk Trigger

public-facing output risk

client-facing output risk

paid traffic risk

compliance risk

financial decision risk

developer handoff risk

private client context risk

Audit Cadence

MWMS uses three audit cadences.

Light Monthly Check

Purpose:

Catch obvious drift early.

Checks:

duplicate files

wrong folders

stale drafts

retired language resurfacing

repeated user corrections

skills producing wrong formats

outputs sounding generic

missing review status

Quarterly Full Audit

Purpose:

Review active systems properly.

Checks:

context accuracy

skill usefulness

voice consistency

offer accuracy

customer language freshness

proof status

folder governance

AI Employee alignment

Brain routing accuracy

human review requirements

Kaizen learning capture

Event-Based Audit

Purpose:

Review immediately after a major change.

Triggered by:

new offer

new client

new launch

campaign result

major user correction

new compliance concern

major course absorption

major Brain update

new AI Employee creation

Audit Objects

MWMS may audit different objects depending on need.

Context Library Audit

Checks whether the context library remains accurate, current, specific, and source-grounded.

Skill Audit

Checks whether a reusable AI skill still performs the correct procedure.

Voice Audit

Checks whether AI output still matches the approved voice.

Offer Audit

Checks whether AI output still matches the current offer.

Proof Audit

Checks whether proof remains approved, current, and safe.

Customer Language Audit

Checks whether customer wording and objections remain current.

Folder Audit

Checks whether context files are stored correctly.

AI Employee Audit

Checks whether the AI Employee role, capability, context, skill, permission, and output behaviour remain aligned.

Client Brain Audit

Checks whether a future AIBS client system remains accurate, isolated, permissioned, and useful.

Standard Audit Workflow

MWMS uses the following audit workflow.

Step 1: Identify Audit Object

Define what is being audited.

Examples:

Offer Context Library

Voice Architecture

Course Absorption Skill

Client Report Skill

Content Brain output

AIBS client library

Step 2: Identify Current Source Of Truth

Locate the current approved source files.

Do not rely only on memory.

Step 3: Review Recent Outputs

Check recent outputs produced from the Brain, skill, or library.

Examples:

ads

emails

reports

content

scripts

page output

client deliverables

developer handoffs

Step 4: Compare Output Against Source Truth

Check whether outputs still match approved context and standards.

Step 5: Identify Drift Signals

Look for stale, generic, inaccurate, duplicated, unsupported, unsafe, or inconsistent behaviour.

Step 6: Classify Drift Severity

Classify as:

No Drift

Minor Drift

Moderate Drift

Major Drift

Critical Drift

Step 7: Decide Audit Outcome

Choose a clear outcome.

No audit should end in uncertainty.

Step 8: Apply Correction Or Route

Depending on the outcome, refresh, split, merge, park, retire, escalate, or update.

Step 9: Record Learning

Add useful lessons to the appropriate log, registry, context file, or skill update.

Step 10: Schedule Next Review

Set next review timing if the object remains active.

Drift Severity Levels

No Drift

The object remains current and useful.

Action:

keep active

next normal review

Minor Drift

Small corrections needed.

Examples:

one outdated phrase

minor voice mismatch

small format issue

Action:

refresh relevant section

keep active

Moderate Drift

Repeated issues appear.

Examples:

frequent user correction

missing new standard

outdated objections

wrong output format recurring

Action:

update skill or context

review related outputs

Major Drift

System behaviour is unreliable.

Examples:

wrong offer promise

stale buyer profile

incorrect proof

wrong Brain routing

conflicting skills

Action:

pause use

repair before further high-value output

human review required

Critical Drift

Output could create serious risk.

Examples:

client context leakage

unsupported claims in ads

developer instructions from stale state

financial recommendation from outdated assumptions

compliance-sensitive false proof

Action:

stop use

escalate to HeadOffice

human review required

possible retirement or rebuild

Audit Outcomes

Every audit must end with one or more clear outcomes.

Current

Object remains valid.

Refresh

Update part of the object.

Split

Divide a broad object into smaller clearer objects.

Merge

Combine overlapping objects.

Park

Keep for later but remove from active use.

Retire

Remove from active use permanently.

Escalate

Send to HeadOffice or relevant Brain for review.

Rebuild

Object is too broken and needs full reconstruction.

Archive

Move old material out of active context.

Audit Outcome Rules

No audit should end with “maybe.”

If confidence is low, use:

Human Review Required

Park For Review

Escalate To HeadOffice

Do not keep uncertain material active without marking it.

Rewrite Versus Polish Rule

MWMS uses the Rewrite Versus Polish Rule to detect skill and output decay.

If the human reviewer only polishes small details, the skill or context may be healthy.

If the human reviewer must rewrite the structure, logic, voice, source grounding, or output format, the skill or context needs review.

Polishing signals normal refinement.

Rewriting signals system decay.

Examples of polishing:

small wording change

minor tone adjustment

one heading adjustment

minor formatting correction

Examples of rewriting:

wrong structure

wrong page style

wrong Brain logic

wrong offer

wrong buyer

wrong context

wrong output format

missing key rules

invented proof

generic AI filler

Context Library Audit Questions

Ask:

Is this library still active?

Is the source-of-truth location clear?

Are any files duplicated?

Are draft files clearly marked?

Are archived files inactive?

Is Retired Language current?

Is the buyer profile still accurate?

Is the offer profile still accurate?

Is the methodology still accurate?

Are objections still current?

Is proof still approved?

Is customer language still fresh?

Are compliance notes current?

Are skills reading the correct library?

Are outputs still matching the library?

Skill Audit Questions

Ask:

Is the skill still used?

Does the task still repeat?

Does the skill still match the current MWMS process?

Does the trigger still work?

Does it trigger when it should not?

Does it require current context?

Does it define forbidden actions?

Does it define validation?

Does it define handoff?

Does it respect tool permissions?

Does it require human review?

Does it overlap another skill?

Does it conflict with another skill?

Does it need refresh, merge, split, park, or retirement?

Voice Audit Questions

Ask:

Does output still sound like the approved voice?

Is AI over-polishing?

Are banned phrases appearing?

Are founder phrases being preserved where useful?

Is tone consistent across Brains?

Is copy becoming generic?

Is humour, directness, or formality drifting?

Is the voice still appropriate for the buyer?

Is the voice file still current?

Offer Audit Questions

Ask:

Does output still describe the offer accurately?

Is the promise correct?

Are inclusions correct?

Is pricing current where relevant?

Is the buyer still correct?

Is risk reversal current?

Are limitations clear?

Are old bonuses or features appearing?

Are new offer details missing?

Is the methodology still represented correctly?

Proof Audit Questions

Ask:

Is proof approved?

Is proof still current?

Is proof being used accurately?

Are claims supported?

Are testimonials allowed?

Are case studies current?

Are compliance-sensitive claims flagged?

Is AI inventing proof?

Is proof being overstated?

Customer Language Audit Questions

Ask:

Are customer phrases still current?

Have new objections appeared?

Have new anxieties appeared?

Have desired outcomes changed?

Are old phrases overused?

Is AI inventing customer language?

Is customer language being used strategically?

Does copy still reflect buyer reality?

AI Employee Audit Questions

Ask:

Does the AI Employee still have a clear role?

Does it have the right context?

Does it use the right skills?

Does it respect tool boundaries?

Does it produce the correct output?

Does it route output correctly?

Does it know when to escalate?

Does it avoid forbidden actions?

Does it require too much correction?

Does it create useful outcomes?

Audit Record Template

Each audit should be recorded using the following structure.

Audit Name:

Audit Object:

Owning Brain:

Audit Type:

Audit Date:

Reviewer:

Current Status:

Source Of Truth Checked:

Recent Outputs Reviewed:

Drift Signals Found:

Drift Severity:

Audit Outcome:

Required Corrections:

Related Files Or Skills Affected:

Human Review Required:

Next Review Date:

Learning Captured:

Notes:

Minimum Audit Record

For quick audits, use:

Audit Object:

Date:

Status:

Drift Found:

Outcome:

Action Required:

Next Review:

Kaizen Connection

This framework connects directly to the MWMS Kaizen Loop.

Every audit should support:

Reflect

What happened?

Reduce

What can be simplified or removed?

Refine

What should be improved?

Record

What learning should be captured?

AI Brain audit work should not only fix individual files.

It should improve the system’s ability to avoid the same failure next time.

Human Review Rule

Human review is required when an audit identifies:

Major Drift

Critical Drift

client-facing risk

compliance risk

paid traffic risk

financial risk

developer risk

MCR source-of-truth risk

client context leakage risk

unsupported proof

wrong offer promise

human-visible public output risk

AI Employees may identify drift, but high-risk correction requires human review.

Client Brain Audit Rule

Future AIBS client systems must include recurring audit logic.

Client Brain audits should check:

client offer accuracy

client voice accuracy

client context separation

client permissions

client approval rules

client reporting accuracy

client skill performance

client proof usage

client compliance constraints

client data sensitivity

Client AI systems should not be sold as set-and-forget.

They must include maintenance and audit expectations.

Governance Role

HeadOffice owns the MWMS AI Brain Audit And Decay Prevention Framework.

HeadOffice is responsible for:

defining audit rules

setting review cadence

classifying drift severity

ensuring high-risk drift is escalated

ensuring audits produce clear outcomes

preventing stale context from staying active

ensuring skills are refreshed or retired

ensuring client systems remain isolated

ensuring AI Employees improve through Kaizen

ensuring audit findings update relevant files, standards, or skills

Individual Brains may run their own audits, but HeadOffice governs cross-Brain, high-risk, MCR, AI Employee, and future client-system audit behaviour.

Relationship To Other MWMS Standards

This framework supports and must align with:

MWMS Document Structure Standard

MWMS AI Agent Memory And Context Framework

MWMS AI Agent Skill Library Framework

MWMS AI Skill Builder And Audit Protocol

MWMS Offer Context Library Standard

MWMS Context Library Governance And Folder Map Standard

MWMS AI Context Activation And Usage Protocol

MWMS Client IP Excavation Framework

MWMS AI Output Validation Standard

MWMS Messy Input Normalization Framework

MWMS AI Employee Role Card Standard

MWMS AI Employee Capability Stack Framework

MWMS AI Employee Evaluation Scorecard Standard

MWMS Agentic Reporting Standard

MWMS AI Agent Outcome Measurement Framework

MWMS Brain Routing Rule

MWMS MCR Promotion To Brain Protocol

MWMS Page Naming Standard

MWMS Architecture Registry

HeadOffice Kaizen Continuous Improvement Loop

Content Brain VOC Grounded AI Copy Framework

Compliance Brain Claims Risk Framework

AI Business Systems Brain Canon

This framework provides the audit and decay-prevention layer for context, skills, Brains, AI Employees, and future client systems.

Drift Protection

This framework protects MWMS from:

stale context

stale skills

generic AI output

outdated voice

wrong offer information

unsupported proof

old language resurfacing

folder decay

client context leakage

wrong Brain routing

governance erosion

manual corrections never becoming system improvements

AI Employees becoming less useful over time

skills remaining active after they are outdated

context libraries becoming bloated or duplicated

future client AI systems becoming unsafe or inaccurate

Any AI system that has not been reviewed after major change or repeated correction should be treated as a drift risk.

Architectural Intent

The architectural intent of the MWMS AI Brain Audit And Decay Prevention Framework is to keep MWMS intelligent over time.

The value of MWMS is not only in creating Brains, skills, context libraries, and AI Employees.

The value is in keeping them current, useful, aligned, and governed.

The long-term goal is that every active AI Brain, context library, skill, and client system can answer:

Is this still true?

Is this still useful?

Is this still current?

Is this still source-grounded?

Is this still aligned with MWMS standards?

Is this still safe to use?

Is this still producing good outputs?

Does it require polishing or rewriting?

Should it be refreshed, merged, split, parked, retired, or rebuilt?

When MWMS can answer these questions consistently, the ecosystem becomes more resilient, more accurate, more scalable, and safer for future client delivery.

Change Log

v1.0 — Initial Draft

Created the MWMS AI Brain Audit And Decay Prevention Framework as the audit and maintenance framework for preventing decay across MWMS Brains, AI Employees, context libraries, skill systems, offer files, voice systems, customer language banks, proof libraries, folder structures, and future AIBS client systems.

This framework defines decay types, audit triggers, audit cadence, audit objects, audit workflow, drift severity levels, audit outcomes, rewrite-versus-polish rule, audit questions, audit records, Kaizen connection, human review requirements, client Brain audit rules, governance role, drift protection, and architectural intent.

Change Impact Declaration

Pages Created:

MWMS AI Brain Audit And Decay Prevention Framework

Pages Updated:

None

Pages Deprecated:

None

Registries Requiring Update:

MWMS Architecture Registry

HeadOffice Page Registry

AI Business Systems Brain Page Registry

Content Brain Page Registry

Offer Brain Page Registry

Canon Version Update Required:

No

Change Log Entry Required:

Yes

Employee Impact Check

Employees impacted:

HeadOffice Manager Employee

AI Manager

AI Employee Router

Context Library Builder

Client IP Excavator

Skill Auditor

Offer Strategist Employee

Content Planner Employee

Creative Strategist Employee

Sales Strategist Employee

Research Analyst Employee

AI Business Systems Architect Employee

Required behaviour updates:

AI Employees must treat Brains, context libraries, skills, and client systems as living systems that require review.

AI Employees must flag repeated human rewriting as a decay signal.

AI Employees must identify stale context, stale skills, old language, unsupported proof, wrong offer details, and client context leakage risks.

AI Employees must route moderate, major, and critical drift to the correct review pathway.

AI Employees must capture useful audit findings as Kaizen learning and update relevant context, skill, or governance records after approval.

END MWMS AI BRAIN AUDIT AND DECAY PREVENTION FRAMEWORK v1.0