Skip to content

Free readiness scorecard

Can your semantic model support defensible AI answers?

Score one priority Power BI model against 21 checks. You will see the complete result without entering contact information. Allow 8–12 minutes if you can consult your model documentation.

Directional A self-reported diagnostic, not certification or an audit opinion.

Private by design Answers stay in this tab and are not sent to Refinity.

Evidence-led “Yes” means the control is documented and tested—not merely intended.

Assessment progress

0 of 21 answered

01 / 07

Model structure

A deliberate, navigable model with explicit relationships and a constrained surface for AI.

1.Are table relationships intentional, valid, and tested for the priority use case?

Evidence: Relationship review, ambiguity check, representative query tests

2.Is the user-facing model surface limited to relevant, well-named fields and measures?

Evidence: Hidden technical fields, curated perspectives or AI data schema

3.Are consequential calculations implemented as explicit measures rather than implicit aggregation?Critical control

Evidence: Measure inventory and visual/query dependency check

02 / 07

Business terminology

Names, definitions, synonyms, and ownership that match how the target users ask questions.

1.Do business-facing tables, columns, and measures use language the target users recognize?

Evidence: Model labels compared with real user questions

2.Are important terms defined with scope, exclusions, and an accountable owner?

Evidence: Metric glossary or definition register with owners

3.Are synonyms and ambiguous phrases mapped to the intended business concepts?

Evidence: Vocabulary map covering representative phrasing

03 / 07

Measures and data foundations

Trusted explicit measures with traceable source, grain, calendar, filters, and ownership.

1.Is there one authoritative measure for each priority business concept?Critical control

Evidence: Approved measure list and duplicate/near-duplicate review

2.Are source, grain, calendar, filters, and ownership documented for priority measures?

Evidence: Metric answer contract or semantic specification

3.Are data quality, refresh timing, and known coverage limits tested and visible?

Evidence: Freshness checks, reconciliation evidence, limitation notes

04 / 07

AI context and instructions

A focused AI data schema plus instructions that resolve business language without contradicting the model.

1.Is the AI data schema focused on the tables, fields, and measures needed for this use case?

Evidence: Purpose-built AI-visible schema reviewed against question set

2.Do AI instructions clarify business rules without duplicating or contradicting model logic?

Evidence: Instruction review with conflict and precedence tests

3.Are AI schema and instruction changes versioned, reviewed, and tied to evaluation results?

Evidence: Git history, review record, evaluation run

05 / 07

Verified answers

Authoritative responses for high-value questions, with known scope and maintenance ownership.

1.Do the highest-consequence recurring questions have an approved answer path?

Evidence: Verified-answer inventory mapped to business questions

2.Can each approved answer be traced to its measure, filters, time context, and evidence?

Evidence: Answer contract and reproducible reference result

3.Does each verified answer have scope limits, an owner, and a review trigger?

Evidence: Ownership and lifecycle fields in the answer register

06 / 07

Security and governance

Permissions, sensitive fields, ownership, and change controls tested through the AI experience.

1.Have representative user roles been tested through the intended AI experience?Critical control

Evidence: Role-by-question permission tests, including denied cases

2.Are sensitive fields excluded or governed in every AI-visible surface?

Evidence: Sensitive-field inventory and AI schema review

3.Are model, metric, security, and AI-context owners named with approval responsibilities?

Evidence: RACI or operating-control register

07 / 07

Evaluation and lifecycle controls

A repeatable golden-question set, regression checks, owners, and release decisions.

1.Is there a representative golden-question set with expected answers and evidence?Critical control

Evidence: Versioned questions, expected values, tolerances, and citations

2.Are model and AI-context changes regression-tested before release?

Evidence: Pre-release comparison and documented disposition of failures

3.Is there a process to review failures, update tests, and re-approve changed answers?

Evidence: Failure triage, ownership, cadence, and change record

Your result

Answer 21 more questions to calculate your readiness.

The complete result includes seven domain scores, critical blockers, evidence gaps, and three actions. No contact details are required.