What frameworks do AI readiness assessments use?

Ask which publisher, version and method sit behind a readiness result. A maturity model, an adoption guide and a task diagnosis can produce different outputs. Before comparing scores, check what each assessment measures and whether it answers your actual decision about the business.

What does Microsoft’s current guidance cover?

Microsoft’s current AI adoption page frames use cases, organisational foundations and the choice of solutions. It is adoption guidance, rather than a single universal readiness score. Its current framing should not be assumed identical to an older description of phases.

For an assessment based on that guidance, ask which parts were used, which evidence was collected and how recommendations follow from it. Naming a publisher does not show that the entire method was applied or that the publisher endorses your result.

Do not read a distinction between generative and nongenerative models as a guarantee that an output will be consistent or correct. Define the task, acceptable errors, tests and responsibility for failures in the actual proposal. A category name cannot substitute for those decisions.

What can you verify about Gartner’s model?

Gartner’s June 2026 assessment description refers to capability categories, priority gaps and recommended next steps. A short public summary is not the full scoring method; ask which edition and licensed material a provider uses.

This also shows why it is too broad to say a maturity assessment can only return a score. Check the actual deliverables: they may include recommendations or a roadmap. Compare those outputs with your needs instead of deciding from the word “maturity” alone.

A production label should not be read as permission to leave a system unmonitored. Ask who owns the process, how performance is checked and what happens when it fails. The result of an assessment does not remove operational responsibility.

What does Google Cloud’s historical framework measure?

Google Cloud’s 2020 AI Adoption Framework groups six themes across three maturity phases. Its themes cover leadership, learning, access, scaling, automation and security. This is a dated publisher framework, not proof that every current assessment uses the same grid.

If an offer uses that model, ask which business unit and evidence each judgement covers. A broad organisational view can expose a gap in skills or data access, while a particular task may need more detailed observation before you can choose a first build.

Do not treat similar labels from different publishers as an official conversion table. A crosswalk is an interpretation unless the sources establish the equivalence. State the definitions used and check whether the two assessments even measure the same object before comparing their levels.

What should sit behind a readiness score?

The model’s name, publisher, version and the part actually applied.
The scope: business unit, process, people and systems included.
The evidence, including which answers are self-reported and which were checked.
The scoring or judgement method and its limitations.
The output: score, findings, recommended tasks or roadmap, with responsibility for the next step.

A long questionnaire is not automatically more useful than a focused diagnostic, and a short interview cannot claim evidence it did not collect. Match the method to the question. You may need an organisation-wide assessment, a closer look at one task or both, with the handover between them made explicit.

ErgoStrat’s AI Opportunity Map examines five work areas and traces a real task to a measured baseline. It returns named tasks ordered by the hours they cost today. Compare that defined output with the assessment you already have, rather than assuming a model name alone tells you what to build first.

Questions

Which model is the right one?

Choose the question first. Ask whether the method addresses organisational capability, adoption planning or a specific operational task. Then check evidence, version and output. There is no reason to treat different publisher labels as a single score scale.

Does a successful pilot prove operational readiness?

No. Check ownership, data, access, testing, monitoring and response arrangements for continued use. A demonstration and an operating process have different evidence requirements. The assessment should say which of those it actually examined.

Does a public description give access to the full method?

Not necessarily. Ask what material you can inspect and which scoring or licensing terms apply. A public summary can describe a tool without publishing all its questions, weights or interpretation rules. Do not present a secondary outline as a direct reading of the complete method.

Sources

See which tasks rise to the top, without a scale or level

The AI Opportunity Map orders tasks by the hours they cost you today; for the first three it gives a proposed tool, requirements, an owner and a build-effort range in hours, without a build price. It rests on none of the three published readiness models and hands back a list of named tasks, not a number on a scale.

Get your Opportunity Map

or write to contact@ergostrat.com