AI readiness assessment: 16 questions about evidence and responsibility

An AI readiness assessment can start with one intended use: what evidence is available, what is missing and who can provide it. A total score cannot predict staff adoption or replace production acceptance and risk decisions.

The following editorial worksheet groups 16 questions into data, workflow, team and risk. It has no validated weights, industry norms or deployment score. Add or remove questions for the task and record why.

Establish the decision being made

Exploring a use, running a restricted test and launching in production require different conditions. Define the scope being considered: users, eligible data, permission to change external records, review and stop conditions. The evidence then has a specific decision to support.

The NIST AI RMF organizes risk work into GOVERN, MAP, MEASURE and MANAGE. It states that its actions are neither a fixed checklist nor a necessarily ordered sequence. Organizations apply and revisit them in context. The four question groups below do not map directly to NIST's four functions and are not an official certification form.

Sixteen questions and inspectable materials

The required detail depends on the intended use. Tests may use approved synthetic data; production use may require further permissions and responsibility checks. Manual export is not automatically a failure, and recent updates alone cannot establish that data fits a task.

The GAO AI accountability framework covers governance, data, performance and monitoring, with questions for organizations and assessors to consider. It supports ongoing examination without assigning weights or a pass mark to this 16-question worksheet.

Turn unknowns into assigned work

For each question, record evidence available, evidence needed, or not applicable with a reason. Add the responsible person, the material to obtain and the decision that depends on it. These are collaboration states, not universal deployment readiness levels.

For example, an internal knowledge assistant may not yet have a confirmed limit for stopping document access after a user leaves a group. The work item could ask identity and knowledge-system owners to establish synchronization conditions, reproduce the event with approved test accounts and record document, citation and cache behavior. They can then assess those results against the deployment's requirements. This is a hypothetical example, not a completed test.

If scoring helps prioritize work, document its definitions, evidence and limitations. A total cannot offset missing data-use permission, incomplete required acceptance work or unassigned operations responsibilities. Whether a gap blocks a particular step depends on the actual use and the accountable decision maker.

Decide the next step from acceptance evidence

An assessment can lead to gathering materials, narrowing a test, proceeding under confirmed conditions or stopping the current proposal. Record the reasons and unresolved items. Recheck when versions, data, usage or risks change after launch; an initial assessment is not permanent proof of readiness.

To compare delivery arrangements, read build, buy and FDE options. For receiving and acceptance conditions, see the AI project handover guide. In a scope discussion, this worksheet can explain what is already known and where further assessment is needed.