Assess whether a workflow is ready for AI
Check data access, process stability, review capacity and success criteria before starting an AI integration.
AI readiness is a property of a particular workflow, not a badge for the whole company. A business may be ready to draft routine replies while being unready to automate pricing decisions. Assess a concrete use case against its information, permissions, failure consequences and ability to review outputs.
Check whether the underlying work is understood
Ask the process owner to describe a good result and show representative examples. If different staff members disagree about the correct outcome, document the disagreement before testing models. AI cannot resolve an undefined service policy merely by producing a confident answer. A written decision rule may be the first deliverable.
Look for recurring work with accessible inputs and a meaningful completion state. Review whether the process is changing because of a new product, software migration or staffing change. A pilot can still proceed during change, but its scope and evaluation examples must identify which version of the process they represent.
Inspect access and review capacity
List the records the workflow needs, who owns them and which actions the integration would perform. Start with the smallest useful access set. A successful demonstration using one employee’s personal account does not establish that the business has suitable shared access, continuity or permission to process every document in production.
Confirm that somebody can review uncertain cases during actual operating hours. Estimate that workload using sample outputs rather than assuming review will be effortless. If a proposed system needs a specialist to check every result and that specialist is unavailable, the review queue becomes a new bottleneck instead of a safeguard.
Make the readiness decision explicit
Use three decisions: ready for a bounded pilot, ready after named prerequisites, or unsuitable for the proposed level of automation. For each prerequisite, assign an owner and completion evidence. Examples include agreeing a category list, removing obsolete documents or creating a test account with limited permissions.
Prepare a small evaluation set and a manual fallback before selecting an operating model. NIST’s AI risk framework provides a broader reference for managing AI risks; your practical readiness record should remain specific to the task. Do not turn a checklist score into a guarantee of accuracy or commercial value.
Practical checklist
- Write the intended result in business terms.
- Check access using a limited test identity.
- Confirm reviewer availability and fallback ownership.
- Record blockers with owners and completion evidence.
Illustrative setup: a shared sales inbox
An equipment supplier wants automatic replies. Its product catalogue is current, but delivery promises live in individual spreadsheets. The readiness decision permits classification and draft preparation using approved product information. Delivery commitments remain manual until an authoritative source and an owner for updates are established.
Common questions
Can a small company be ready without a data team?
Yes, for a narrow workflow with clear records and an accountable owner. The required controls depend on the proposed actions, not simply on company size.
Does readiness mean the pilot will save money?
No. Readiness means the prerequisites are sufficiently understood to test the idea. The pilot still needs to measure effort, errors and operating cost against the current process.
Further reading
Start with your actual workflow.
Turn the useful parts of this guide into a focused project brief.
Shape your project