AI can help you get past a blank page. In instructional design, the harder work still belongs to you: deciding what learners need to do, checking whether the practice feels real, and making sure the feedback teaches the right lesson. Here's a small workflow you can try with one activity.
Start with a decision, not a prompt
Pick one learning objective. Rewrite it as a decision someone might make on the job. For example, if the objective is “identify when to escalate a customer issue,” the practice moment could be a short customer message with two plausible next steps. Use a fictional or approved scenario for this exercise.
The point is to define success before asking a tool for ideas. What would a good decision look like? What common mistake should the activity help a learner notice?
Ask AI for options
Try a prompt like this:
I am designing a short practice activity for adult learners. The goal is [objective]. Create three fictional workplace situations in which a learner must make a decision. For each, give two plausible choices and draft feedback explaining the consequence of each choice. Avoid facts or policies you cannot verify.
Treat the output as raw material. Choose one situation that resembles a real decision without exposing a real person's data. Rewrite the details in your own language. If your organization has AI or data-handling rules, follow those rules before entering any material into a tool. Product data protections vary by plan and settings; check the tool's current policy rather than assuming all accounts work alike. OpenAI, for example, states that its Business, Enterprise, Edu, and API offerings do not use organization data for model training by default.
Make the choices believable
Read the three situations and keep the one that resembles a decision your learners might really face. Then edit it in your own words. AI may invent a policy, suggest a step your team does not take, or make one choice so silly that nobody would select it. Two plausible options give learners something to reason through.
In the fictional escalation example, one choice might be to gather a missing detail and the other to escalate immediately. Which is right depends on your actual procedure. The activity should provide enough context for a learner to decide without guessing what the author wants. If that context is missing, add it before polishing the wording.
Do the designer's review
Read the activity as a learner would. Can they make a meaningful choice without guessing what the author wants? Are both choices believable? Does the feedback explain why the better choice works and what the other choice might cause? Does every claim match your source material?
If the AI invents a policy, number, or outcome, remove it. If the scenario is too easy, make the alternatives more realistic. If the feedback only says “correct,” add a reason the learner can use later.
Check each claim
Go through the activity with the answer key hidden. Notice where you have to assume a fact that the learner cannot see. Then check every substantive claim against your approved source material: policy steps, product behavior, numbers, and technical terms. If you cannot verify a detail, remove it or label the example as fictional.
Ask whether the feedback gives a reason the learner could use in a different situation. “Correct” can feel encouraging, but it leaves out the useful part. For a less helpful choice, explain which signal was missed and what to look for next time. Keep the feedback brief enough to read comfortably.
Run a ten-minute practice test
Give one colleague the activity without an explanation. Ask them to think aloud while they choose. Notice where the wording, context, or feedback loses them. Usability testing guidance from Digital.gov recommends observing people as they attempt a task and inviting them to think aloud. This tiny check does not prove the activity works for everyone, but it can reveal one change worth making before you share it more widely.
A simple way to close the test is to ask, “What information were you looking for?” and “What did you expect that choice to do?” Revise one point of confusion, perhaps a label, a missing fact, or a feedback sentence. One colleague's attempt is a first observation, not proof that the activity works for everyone.
For today's exercise, take a recall question you already use. Turn it into one short situation with two believable actions. Draft feedback for both, then ask someone to try it without your explanation.
Your finished activity can be small: one situation, two choices, and feedback that teaches. Try the workflow on a question you already have. If you want to turn that first idea into something playable, explore Make a Game.