Skip to content
FREE PLANNING TOOL

AI Workflow Pilot Readiness Assessment

Before an AI workflow goes live, a team needs more than a promising demo. Use eight checks to expose missing ownership, baseline measures, evaluation, and fallback evidence—then start a smaller, more defensible pilot.

No signupBrowser-only answersNot a certification
EIGHT EVIDENCE CHECKS

Check a proposed pilot

Choose the best current answer. This is a planning aid, not a certification, legal assessment, or risk guarantee.

1. Is one workflow and its AI handoff clearly scoped?

Name the start, end, inputs, and human handoff.

2. Do you have a baseline to compare against?

Use time, quality, cost, error rate, or another current outcome.

3. Has data access and out-of-scope data been reviewed?

Confirm what data can enter the pilot and what cannot.

4. Is a human owner accountable for decisions and escalation?

A tool is not an accountable owner.

5. Is there a quality and safety evaluation plan before launch?

Decide how you will inspect outputs and record failures.

6. Can the team stop the pilot and return to a safe fallback?

Document the manual process or rollback path.

7. Have people affected by the workflow been included?

Include operators, reviewers, and downstream users.

8. Is one measurable success metric defined?

Use a measurable target and a review date, not a vague promise.

HOW TO USE IT

Score evidence, not enthusiasm

Each check has equal weight: yes is one point, partly is half a point, and no is zero. The score is a transparent conversation starter, not a prediction of value or safety.

“Ready to pilot” requires evidence for all eight checks. A missing human owner and missing fallback together produce “Not ready,” because neither an automated tool nor a vague plan can take responsibility when a pilot needs to stop.

A PRACTICAL FLOW

Turn the result into a pilot plan

  1. 1. Scope one workflow. Pick a bounded process and state which decision remains human.
  2. 2. Record a baseline. Measure the current outcome before asking whether the pilot improved it.
  3. 3. Set a review date. Define the metric, quality sample, escalation owner, and fallback before use.
  4. 4. Pilot, inspect, and decide. Compare observed results with the baseline; do not infer broad readiness from a demo.
SCOPE AND LIMITS

What this assessment does not decide

This tool does not determine legal compliance, security approval, model suitability, or whether a pilot will deliver financial value. It cannot replace domain review, affected-worker input, security assessment, or monitoring. Treat a partial answer as a prompt to gather evidence rather than a reason to hide uncertainty.

The checks are an AQ Score planning model informed by the voluntary NIST AI Risk Management Framework, which helps organizations incorporate trustworthiness considerations into AI design, development, use, and evaluation. AQ Score is not affiliated with NIST, and this result is not a NIST assessment.

Plan the economic case after the readiness check

Once a pilot has ownership and measurement evidence, model its time and cost assumptions separately. An ROI estimate should not substitute for quality, safety, or worker input.