01 / Understand the opportunity
Start with the outcome, the people affected, and the way work happens today. Examine customer needs, operating constraints, existing measures, systems, and data. Capture uncertainty rather than hiding it inside a business case.
Working output
An opportunity map with a defined business question, initial value hypothesis, baseline, dependencies, and evidence gaps.
Decision to proceed
Is this opportunity important enough to investigate, with an accountable sponsor and access to the people and information needed?
02 / Redesign the work
Describe the future workflow and the capability it enables. Make decisions, exceptions, handoffs, and human authority explicit. Compare the proposed change with simpler process or software improvements.
Working output
A future-state workflow, responsibility map, and assessment of technical delivery options.
Decision to proceed
Does the design address the business need, and are its data, integration, control, and adoption requirements feasible?
03 / Validate before expanding use
Define representative examples, difficult cases, and acceptance criteria. Test outputs against reliable evidence, business rules, or observed feedback wherever possible. Specify when the system should stop, escalate, or request a human decision.
Working output
An evaluation plan that connects test cases and failure severity to business and technical acceptance criteria.
Decision to proceed
Is there enough evidence for the intended use, and are the remaining limitations understood by the people accountable for accepting them?
04 / Prepare implementation and adoption
Translate the design into an implementation plan with agreed owners. Address access, integration, security, user preparation, support, monitoring, and fallback procedures together. Confirm delivery responsibilities before work enters live operations.
Working output
A delivery and readiness plan covering responsibilities, prerequisites, acceptance, user adoption, and ongoing operation.
Decision to proceed
Have the business and technical owners accepted the scope, controls, operational responsibilities, and conditions for release?
05 / Measure and decide what comes next
Compare actual performance with the baseline and the original value hypothesis. Look at business results, user behavior, operating costs, and failure patterns together. Use that evidence to improve, expand, pause, or retire the capability.
Working output
A measurement framework and decision record that separate observed results from estimates and future targets.
Decision to proceed
Is the capability producing useful value within acceptable limits, and what evidence supports further investment?
A focused starting point
An initial engagement can focus on a business opportunity or a workflow with an unresolved decision. Useful inputs include a sponsor, a process owner, current operating information, and access to relevant technical stakeholders. The scope should state the questions to answer, the outputs required, and the decisions those outputs will support.
Questions before
the next step.
Who needs to participate?
Usually a business sponsor, the person accountable for the workflow, people who do the work, and the relevant systems or data owners. Security, legal, and other specialists participate where the intended use requires their judgment. The engagement defines the actual participants and time commitment.
How are scope, cost, and timing determined?
They depend on the decisions to resolve, the number of workflows and systems involved, data access, integration complexity, evaluation requirements, and client availability. These are agreed for the engagement; no standard price or duration is assumed.
How do you address security and sensitive information?
Identify the information involved, approved environments, access requirements, and applicable client controls before using it. The delivery plan should specify data handling, permissions, logging, retention, and review responsibilities. Specific platform or compliance claims require verification for the chosen design.
What makes an AI-enabled workflow ready to use?
Readiness means meeting agreed criteria for a defined use, with representative test evidence, understood limitations, operational controls, and named owners. A successful demonstration alone does not establish readiness. Human review must have a clear purpose and enough context to be effective.
What does the client own?
The agreement should make ownership of data, deliverables, code, configurations, and third-party services explicit. It should also name who will maintain and operate the capability after the engagement. Those terms are resolved before the relevant work begins.
When is AI the wrong answer?
When a simpler change meets the need, the necessary data or access is unavailable, the value does not justify the operating burden, or acceptable performance cannot be demonstrated for the intended use. Deciding not to proceed can be a useful result of the work.