Buyer Simulation
Company-blind by default. Neutral customer scenarios examine discovery, consideration, appropriate recommendation and recommendation boundaries.
Methodology
How maesee measures AI Distribution Readiness
Evidence before opinion. maesee observes AI behaviour, inspects published business information and applies defined scoring rules. Every result needs an interpretation grounded in its test conditions.
Methodology record
Framework v1.1
- Headline metric
- AI Distribution Score /100
- Dimensions
- Five × /20
- Sources of evidence
- Two sources of evidence
- Scoring
- Defined rules
- Confidence
- Separate from score
- Action
- AI Routing /10 + Agent /10
Evidence before opinion
Two sources of evidence support the diagnostic.
What AI says
Responses from defined scenarios, named models and stated test channels.
What the business makes available
Direct evidence from your offer, suitability information and action routes.
- 01Structured evidence
- 02Validation
- 03Scoring rules
- 04Report
AI may assist evidence extraction. Defined rules calculate the score. Confidence stays separate.
The testing instrument
Different questions, a structured purpose.
Business context is confirmed before testing. Versioned testing instruments establish the diagnostic purpose and evidence to observe, without handing scoring decisions to the model.
Diagnostic Core
Controlled probes examine the business’s offer, customer fit, factual representation, uncertainty and the next action.
Competitive Intelligence
Deliberate comparative testing adds context in the full Assessment. Confirmed competitors and observed alternatives are distinguished; findings are not a market ranking.
Named models and test channels
Know what was actually tested.
Model identity is part of the record.
A report states the named primary tested model and any comparison models, alongside the extraction and analysis models used. A model’s API response is not assumed to be identical to its consumer product.
We do not imply every model was tested, or blend model results into an unsupported universal claim.
Channels stay explicit.
A parametric test uses the prompt and the model’s own knowledge, without supplied live retrieval. Retrieval-grounded testing and controlled agent execution are distinct channels for future use where explicitly supported.
Direct inspection of a business source is evidence of the information available; it does not turn a parametric model test into live web retrieval.
From response to report
Response → structured evidence → validation → rule-based scoring → report
Unknown is not failure
Technical gaps, missing responses and unavailable observations are stated. An incomplete metric is shown as unavailable, rather than presenting known observations as a complete aggregate. Confidence describes evidence coverage separately.
Audit provenance
Results are tied to a confirmed profile, date, named model, test channel, evidence snapshot and versioned methodology. Evidence references support traceability without publishing internal scoring instructions.
Interpretation and limitations
Diagnostic indicators, within stated conditions.
Behaviour can change.
Findings describe what was observed in the stated scenarios. They are not a universal AI ranking, a guarantee of recommendations or commercial results, or a prediction of every future interaction.
Agent Readiness measures observed capability. It does not guarantee autonomous transaction completion.
Methodology transparency
We explain the framework, sources of evidence, score structure, confidence and limitations. Full prompt packs, exact formulas, thresholds and internal extraction instructions remain protected.
Human review may be used where it improves diagnostic integrity, including ambiguous source evidence.
Read the public product facts