16 Conclusion: Make the Next Decision

A useful pilot can tell you to change course.

16.1 What you should have now

You began with a capability, not a promise that AI would improve everything. The chapters helped you specify permitted data, check a response, design practice, evaluate feedback and choose an appropriate independent demonstration.

Your resulting pilot should contain a task, source material, rubric, worked example, access route and decision about what remains untested. The supplied PDP-2026.1 pack demonstrates those components. It is an original teaching example, not evidence that a real cohort achieved the proposed outcomes.

You do not need to adopt every technique. Choose the method that serves the capability and that you can support responsibly.

16.2 Review the design before launch

Read your student instructions as someone unfamiliar with the activity. Can they tell what to do, which help is permitted, what to submit and how the work will be judged? Can they complete an equivalent route without a paid subscription or personal disclosure?

Then inspect the evidence chain. Does the recommendation cite a supplied fact? Does the feedback apply the actual rubric? Does the independent task require a changed decision, rather than a renamed copy of the practice case?

If those links are missing, revise the design before assigning marks. A polished handout is not a substitute for them.

16.3 Choose one bounded next step

You may be ready to test the materials with a colleague, run an ungraded tutorial or revise an existing assessment through your institution’s process. You may instead need permission, better source material or an accessible alternative. Record that need rather than treating hesitation as failure.

Use the pilot decision sheet. Name the outcome you will inspect, the effort you can support and the condition that would make you stop or change the activity.

After an actual pilot, report what happened without inventing success: who could participate, what the independent work showed, which errors persisted, what support was needed and what remains unknown. A small pilot cannot establish long-term professional capability or general effectiveness across disciplines.

16.4 Keep the method open to correction

Students should be able to challenge AI feedback with evidence. Colleagues should be able to challenge the teaching design in the same way. Share the scenario, criteria and limitations, not only the attractive outputs.

The free edition and downloadable resources are intended to make that scrutiny and adaptation possible. The printed edition offers another way to read and support the work; neither format removes the need to check local policy, professional requirements and learner needs.

16.5 Final activity: a decision note

Allow ten minutes; no AI required. Write five sentences:

The capability my activity targets.
The evidence that will demonstrate it.
The assistance and data use I permit.
The access or assurance limitation I still need to address.
My next action, owner and review date.

A defensible decision may be to proceed, revise, pause or stop. What matters is that the decision follows from the learning purpose and evidence, not enthusiasm for the tool or confidence in its prose.

# Conclusion: Make the Next Decision {#sec-conclusion} > A useful pilot can tell you to change course. ## What you should have now You began with a capability, not a promise that AI would improve everything. The chapters helped you specify permitted data, check a response, design practice, evaluate feedback and choose an appropriate independent demonstration. Your resulting pilot should contain a task, source material, rubric, worked example, access route and decision about what remains untested. The supplied PDP-2026.1 pack demonstrates those components. It is an original teaching example, not evidence that a real cohort achieved the proposed outcomes. You do not need to adopt every technique. Choose the method that serves the capability and that you can support responsibly. ## Review the design before launch Read your student instructions as someone unfamiliar with the activity. Can they tell what to do, which help is permitted, what to submit and how the work will be judged? Can they complete an equivalent route without a paid subscription or personal disclosure? Then inspect the evidence chain. Does the recommendation cite a supplied fact? Does the feedback apply the actual rubric? Does the independent task require a changed decision, rather than a renamed copy of the practice case? If those links are missing, revise the design before assigning marks. A polished handout is not a substitute for them. ## Choose one bounded next step You may be ready to test the materials with a colleague, run an ungraded tutorial or revise an existing assessment through your institution's process. You may instead need permission, better source material or an accessible alternative. Record that need rather than treating hesitation as failure. Use the [pilot decision sheet](../appendices/pilot-pack.qmd). Name the outcome you will inspect, the effort you can support and the condition that would make you stop or change the activity. After an actual pilot, report what happened without inventing success: who could participate, what the independent work showed, which errors persisted, what support was needed and what remains unknown. A small pilot cannot establish long-term professional capability or general effectiveness across disciplines. ## Keep the method open to correction Students should be able to challenge AI feedback with evidence. Colleagues should be able to challenge the teaching design in the same way. Share the scenario, criteria and limitations, not only the attractive outputs. The free edition and downloadable resources are intended to make that scrutiny and adaptation possible. The printed edition offers another way to read and support the work; neither format removes the need to check local policy, professional requirements and learner needs. ## Final activity: a decision note **Allow ten minutes; no AI required.** Write five sentences: 1. The capability my activity targets. 2. The evidence that will demonstrate it. 3. The assistance and data use I permit. 4. The access or assurance limitation I still need to address. 5. My next action, owner and review date. A defensible decision may be to proceed, revise, pause or stop. What matters is that the decision follows from the learning purpose and evidence, not enthusiasm for the tool or confidence in its prose.