The practical starting point
AI readiness is not a technology shopping exercise. It is a check on whether the business has the process, data, ownership and safeguards needed for AI to help rather than confuse.
Use this checklist before buying tools, building agents or automating customer-facing tasks.
You do not need every item perfect, but weak areas should shape the first project scope.
The 25-point AI readiness checklist
Use these points to identify whether a workflow is ready for AI-assisted improvement.
- A valuable workflow is selected
- The trigger is clear
- The desired output is defined
- The human owner is named
- Exceptions are known
- The fallback route is documented
- Source data is accessible
- Source data is accurate enough
- Sensitive data is identified
- Data minimisation is planned
- Permissions are clear
- Privacy information is reviewed
- A lawful purpose is understood
- Approved knowledge sources exist
- Output quality criteria are defined
- Human review is placed at risk points
- Users know when AI is involved
- Prompt and workflow changes are controlled
- Testing scenarios include edge cases
- Logs can be reviewed
- Failures trigger alerts
- Staff know how to override
- Support ownership is assigned
- Costs can be estimated from volume
- The first release is small enough to learn from
How to score readiness
Score each item as ready, partial or not ready. A high-risk workflow with many partial answers should not be the first AI project.
Look for a contained use case where the system can assist preparation, classification, routing or drafting while a person remains accountable.
Privacy and governance questions
If the workflow uses personal data, assess fairness, transparency, minimisation, access and retention. AI does not remove normal UK data-protection obligations.
For higher-risk processing, consider whether a DPIA is needed and whether meaningful human intervention is part of the process.
What to do after the audit
Pick one workflow with strong readiness and a measurable benefit. Define the test set, success criteria, escalation route and review rhythm.
Treat the first release as a controlled pilot. Expand only after the team trusts the output and understands the limits.
For a governed build, review company-customised AI agents.
Frequently asked questions
It is a structured review of workflows, data, people, risk, privacy and support before using AI in a business process.
No, but the data must be good enough for the use case and the system must handle uncertainty safely.
A bounded, frequent workflow where AI assists preparation or routing and a human can review important outputs.
Audit readiness before implementation
Fekitech can help you assess AI readiness, choose the first suitable workflow and design the controls around it.
