Start with a real workflow
Name the work being done, who does it and who experiences the result. “Use AI in operations” is an ambition, not a decision boundary. A narrower question might be whether staff could draft a response from approved source material while retaining responsibility for the final answer.
Record the current process before proposing a replacement. Include the exceptions, hand-offs and informal checks that keep it functioning.
Establish a baseline before claiming a benefit
Decide what improvement would mean: less rework, shorter handling time, better access to information or fewer missed review steps. Measure the existing process where possible. Separate observed values from estimates and state which costs the estimate excludes.
A demonstration does not establish sustained value. Test against representative work, including difficult cases, and include review time and operating effort in the comparison.
Check data access and permission
Identify the sources the workflow needs and who can authorise their use. Assess quality, currency, access restrictions and whether information can be shared with a chosen service. A retrieval system does not make unreliable source material trustworthy.
Start with approved, non-sensitive examples. Do not place personal, confidential or regulated information into a trial simply because an interface accepts it.
Define the human decision
Write down what the system may suggest, what it may do and what it must leave to a person. Make the reviewer, escalation route and fallback process explicit. A nominal human approval is not enough unless that person has time, information and authority to challenge the output.
For higher-consequence work, the evaluation and approval requirements should reflect the potential harm of an error.
Choose the next evidence-gathering step
The next step may be process clarification, a data-quality assessment or a small evaluation rather than an implementation project. State the question the step will answer, the evidence it will produce and the condition that would stop further investment.
A useful readiness assessment ends with an owned decision. It should make “not yet” as actionable as “proceed”.
Further reading
This is Theory Y’s practical interpretation, not a reproduction of a standard or a claim of certification.
