Oprill field notes · workflow planning
Change one workflow with a clear reason.
Five original planning exercises for deciding where AI might fit into a team’s work. Start with an observable problem, identify the reviewer and agree on the conditions for stopping.
Original educational exercises. This is not a collection of executive interviews, customer results or research findings.
1. Describe a friction point.
Use a fictional team that repeatedly reconstructs the status of a support issue from scattered notes. Before proposing a new tool, write down why the work is difficult: inconsistent status labels, missing owners or conflicting sources may each require a different response.
State one observable outcome, such as a reviewer being able to identify the current owner and unresolved decision from the approved notes. Avoid a broad aspiration like transforming the whole organisation. A bounded problem makes it easier to recognise whether a change helps.
2. Separate the task from the action.
Sketch the work in three stages: collect the permitted context, prepare a draft and decide what happens next. Give each stage an owner role. In the fictional support exercise, drafting an internal status note should not authorise sending a customer message or closing the issue.
Write the boundary next to the proposed task. List the actions that require a separate human decision, including changes to access or shared records. A useful trial makes that boundary visible even when the drafted answer sounds convincing.
3. Build a case that could disprove the idea.
Create a straightforward case, a missing-source case and a contradictory-source case. For the contradictory case, include an old note that says the issue is resolved and a newer note that asks for more investigation. Define the expected reviewer response before running the exercise.
The purpose is to find where the workflow fails, not to assemble a demonstration that always succeeds. Keep the source set small enough to inspect, and use fictional information until the relevant owners approve a workplace evaluation.
4. Include the people doing the review.
Ask the reviewer to record corrections and uncertainty as well as elapsed time. A draft can transfer work rather than remove it. If the reviewer spends most of the exercise reconstructing sources, better source labels may matter more than a different instruction.
Agree on how feedback changes the next trial. Change one part of the workflow at a time, keep the original cases and record why a revision was made. Do not turn a single successful attempt into a claim of consistent performance.
5. Make the next decision reversible.
Write a short decision note with three options: stop, revise or conduct another bounded exercise. State what evidence supports the choice, what remains unknown and who owns the next action. A recommendation to stop is a valid outcome when the task or review burden is not understood.
Use the planning checklist alongside this playbook to identify missing preparation. Checking a box records your own planning judgement only; it does not certify readiness or send information anywhere. Save the standalone workbook if you need a stable copy for discussion.