RESOURCES · Overview · White Paper · Evidence Methodology · Applied Evidence · System Map · Current Thinking · FAQ
mAI Evidence System
Evidence before claims
A transparent methodology for separating demonstrated capability, measured client results, operating targets and claims that require additional validation. Case-study results are measured in live client deployments, with client names withheld at their request. They are not a forecast or guarantee.
Technology Overview · Architecture · Evidence & Governance
Evaluator route: classify the evidence → establish the measurement contract → apply decision authority → review the claim.
What this page is for
mAI is designed around measurable operating decisions. Every result should make the baseline, context, measurement window and limits of the evidence clear. See where each approval gate sits in the platform workflows.
The core rule
- A Named-Company Test Model is not a client result.
- A measured client result is not a universal benchmark.
- An operating target is not a guarantee.
- A compliance capability is not a blanket compliance claim.
01 · Proof ladder
Four levels of proof
Evidence strength increases from capability demonstration to context-specific application, planning target and supported + reviewed claim. Public mAI materials distinguish Named-Company Test Models built from public or non-proprietary context from Real Client Case Studies reporting measured deployment outcomes. Test models are not client deployments or client results.
Capability ≠ outcome · test model ≠ client deployment · real client result ≠ universal benchmark · target ≠ guarantee · validated claim = supported + reviewed for its stated context.
01 · Capability
Demonstration · Capability
Shows how an mAI workflow, decision pattern or architecture can operate. Capability evidence only, not proof of a client relationship or realized business outcome.
02 · Measured
Case Study · Deployment-specific
Reports results measured in a live client deployment over a stated period. The client’s name is withheld at their request. Results remain deployment-specific, are not independently audited and are not a benchmark.
03 · Target
Operating Target · Planning objective
P² planning objective
Approximately 15–20% Productivity improvement and 10–25% Precision improvement over 90 days against an agreed baseline. Actual outcomes vary.
04 · Validated
Validated Claim · Supported + reviewed
A statement supported by an appropriate source, documented measurement or implementation evidence and reviewed for the context in which it is communicated.
Reading the public record: Applied Evidence reports real measured client deployment outcomes within stated measurement boundaries. Named-Company Test Models show how mAI could be configured around public or non-proprietary organizational context without implying a client relationship, endorsement, deployment or realized result. This page sets the classification, baseline, attribution and claims-validation rules that govern both.
02 · Measurement discipline
The mAI Measurement Contract
No baseline, no lift claim. Before interpreting performance, establish the measurement contract. This gives the AI Marketing Brain a reliable decision frame and the AI Marketing OS a consistent Plan → Execute → Measure → Optimize loop.
01 · Define
Baseline + KPI
Define the starting condition, comparison period, KPI definition and calculation method.
02 · Observe
Window + Context
State the observation period and record material changes in budget, channel mix, market, offer, audience or data quality.
03 · Interpret
Confidence + Governance
Separate association, experiment evidence and causal conclusions. Keep Human Command over material strategy, budget, audience, brand and regulated decisions.
03 · Communication standard
How results should be communicated
Public case studies report measured results from live client deployments, with client names withheld at their request. Results should identify operating context and measurement period, avoid implying independent audit, and never be generalized into promises of future performance. Named-Company Test Models use public or non-proprietary context to demonstrate possible system configuration; they are not client deployments or client-result claims.
Regulated environments
mAI can be configured with governance, access, approval and data-handling controls. Compliance depends on the specific deployment, data flows, contracts, infrastructure, organizational controls and applicable legal requirements; it is validated per implementation.
Evidence answers: what is supported? → Governance answers: may the decision proceed?
04 · Decision authority
Human Command scales with decision risk
Automation authority stays inside approved operating boundaries. Human control increases with consequence, uncertainty, brand sensitivity and regulatory exposure. Higher consequence or lower confidence moves the decision toward review, pause or explicit approval.
01 · Lower risk: Bounded automation
Routine, reversible actions may be automated when explicitly approved, monitored and constrained by policy.
02 · Medium risk: Recommend + Review
mAI recommends and explains. A designated human reviews before material execution.
03 · Higher risk: Flag + Pause
Explicit human approval for strategy, audiences, budgets, brand-critical creative, regulated claims and other material decisions.
05 · Outcome model
P² turns results into measurable operating outcomes
Track efficiency and performance separately. A faster operating system is not automatically a better-performing marketing system.
Productivity
time_to_launch · ops_hours_saved · reporting_latency · automation_coverage
Precision
CTR · CVR · CPA/CPL · ROAS/ROMI · LTV · retention · pipeline quality · agreed business KPIs
Measurement gate: targets only become meaningful when tied to a documented baseline, KPI definition, observation window and measurement design.
06 · Review gate
Claim review gate
Advance only when explicit: evidence class · baseline · window · context · confidence · accountable reviewer.
Claim anatomy: what changed · compared with what · over which period · under which conditions · with what level of confidence · approved by whom.
- Classify: Real Client Case Study / Applied Evidence, Named-Company Test Model, operating target or validated claim.
- Source: identify the supporting source, deployment record, model assumptions or public/non-proprietary context appropriate to the class.
- Baseline: for performance claims, state the starting condition, comparison period and KPI calculation.
- Window: define when observation starts and ends when a measured result is involved.
- Context: record material budget, channel, audience, offer, market, data-quality or scenario assumptions.
- Confidence: distinguish illustration, association, attribution, experiment evidence and causal conclusions.
- Human Command: require accountable review for material, brand-sensitive or regulated public claims.
Next inspection
Continue the review path
Inspect the system in Architecture, review measured client results in the Case Studies, and explore named-company test models in the System Map. Read every result with its stated baseline, measurement window and operating context.
STRATEGIC REVIEW · Technology → Architecture → Evidence → Deployment → Leadership
