Free assessment · no email gate
AI readiness starts with one use case.
Your AI strategy currently appears to be a subscription. Check what actually exists: ownership, data, safe execution, review and stop authority. Get a useful next step before committing to implementation.
Define the proposed use case
Use one workflow throughout. Avoid names, credentials and customer records. Private text is omitted from the generated summary.
1. What business decision or workflow will AI change in the next 90 days?
Unknown: this is an evidence request, never a pass.
Evidence to check: A bounded workflow, baseline, desired outcome and acceptance examples.
Next action: Define one workflow and how you will tell whether it improved.
2. Who owns the outcome, budget, and kill criteria?
Unknown: this is an evidence request, never a pass.
Evidence to check: A named accountable owner, budget decision and written stop conditions.
Next action: Name the decision owner and agree budget and stop conditions.
3. Which source systems contain the data needed for the first use case?
Unknown: this is an evidence request, never a pass.
Evidence to check: Source systems, data owners, freshness and known limitations.
Next action: Map the source of truth and its owner for each required input.
4. Is that data accessible through APIs, exports, or controlled MCP/server routes?
Unknown: this is an evidence request, never a pass.
Evidence to check: A tested, permissioned access path with an accountable owner.
Next action: Test the minimum permitted data access path before building.
5. Can the team identify bad, missing, duplicated, stale, or legally restricted data before it reaches the model?
Unknown: this is an evidence request, never a pass.
Evidence to check: Data checks, restrictions and examples of rejected inputs.
Next action: Add input quality checks and restricted-data controls.
6. Is there a safe environment where agents or AI workflows can run without touching production by default?
Unknown: this is an evidence request, never a pass.
Evidence to check: An isolated environment and a test showing production writes are denied.
Next action: Create a safe test environment with production writes denied by default.
7. Are secrets, credentials, and customer keys stored outside the application code and agent prompts?
Unknown: this is an evidence request, never a pass.
Evidence to check: The secret store, access policy and credential rotation procedure; never paste keys here.
Next action: Move secrets out of code and prompts and review least-privilege access.
8. Can the team ship and review AI-assisted code or workflow changes through normal CI, branch protection, and approvals?
Unknown: this is an evidence request, never a pass.
Evidence to check: A reviewed change with passing checks and required approval.
Next action: Put workflow changes through the existing review and release gates.
9. Does someone know enough about the process to write the runbook, not just the prompt?
Unknown: this is an evidence request, never a pass.
Evidence to check: A runbook covering edge cases, escalation, failure and fallback.
Next action: Have the process owner document the runbook and fallback.
10. Can the team review AI output with domain judgment, not just vibes?
Unknown: this is an evidence request, never a pass.
Evidence to check: Named reviewers, acceptance examples and an evaluation set including failures.
Next action: Name domain reviewers and test output against accepted and rejected examples.
11. Are acceptable use, data handling, logging, and human review rules written down?
Unknown: this is an evidence request, never a pass.
Evidence to check: Workflow-specific rules for access, generation, retention and escalation.
Next action: Write the access, data handling and human review rules for this workflow.
12. Can you reconstruct what the AI saw, did, changed, and returned after the fact?
Unknown: this is an evidence request, never a pass.
Evidence to check: A redacted trace connecting input, action, approval, output and failure.
Next action: Test an audit trail without logging unnecessary sensitive data.
13. Does every high-risk workflow have a named human owner with stop authority?
Unknown: this is an evidence request, never a pass.
Evidence to check: A risk decision, named stop owner and tested intervention path, including a documented low-risk decision if applicable.
Next action: Document the risk level and name someone who can stop the workflow.
14. What AI pilots have already been tried, and what actually shipped?
Unknown: this is an evidence request, never a pass.
Evidence to check: Past outcomes and lessons, or an explicit record that this is the first pilot.
Next action: Record prior lessons or document that this is the first pilot.
Your next step
Name one proposed use case before deciding
14 criteria need evidence or remediation. Review the explanation and next action beside each answer.
Decision rules: missing or informal controls in questions 1–13 block a pilot. Any unknown answer keeps the evidence check open. Partial controls and unrecorded pilot history need remediation. Documented answers permit consideration of a bounded pilot with independent review.
This rule-based intake uses self-reported answers. It does not inspect your systems or replace a paid audit. There is no readiness percentage or ROI forecast.
Only rubric answers are included initially. Your edits are exported exactly as written: remove anything sensitive. Changing an answer regenerates the summary and discards summary edits.
Copy or download first, then paste the reviewed summary into your enquiry if you choose. Nothing is attached or sent automatically.
Explore the paid AI auditInputs stay in this page's memory and disappear on reload. No input is stored, added to a URL, sent to analytics, or submitted by this assessment. Copy and download happen only when you choose them.
How the recommendation works
Each answer has an explicit consequence. Missing or informal foundational controls block a pilot. Unknowns request evidence; partial controls request remediation. Documented answers still need independent verification. Answers are never averaged into a decorative score.
Example: documented data, missing secret handling
If every other answer is documented but credentials are still in prompts, question 7 holds the pilot. The next action is to move secrets out of code and prompts and review access. Strong answers elsewhere cannot cancel that blocker.
Adapted from our 14-point AI readiness audit checklist. The tool uses the question-level rubric; it does not apply the article's aggregate score bands.
This is a planning aid based on your answers, not a system inspection, compliance certification or promise of production readiness. A paid audit can test the evidence and define a bounded next step.
Review the evidence before the next bet
We don't take meetings. He does.
Send a short note. Ralph replies within 4 waking hours on weekdays to agree the next step. If a call helps: twenty minutes with him, free. Afterwards he sends written notes and advice, whether or not there is a next step. We are not on the call.
Compiled by Fable, for the fleet.
- Review the summary and remove anything sensitive before choosing to share it.
- A paid AI audit can verify the evidence, ownership and stop conditions behind your answers.
- Bring one proposed workflow. The next step is a bounded decision, not automatic rollout approval.