KINSMC
Pick one request you can actually make
A teammate has twice sent a handoff just before you clock off. You need time to ask questions before taking over, and you want to agree on an earlier handoff. Practising with an AI chatbot can help you hear your opening and try a few replies before the real conversation. The example works best when your request is specific enough for the other person to accept, reject, or amend.
On paper, write the observable event, its effect on your work, and the change you want. “The last two handoffs arrived just before I left” describes something you could verify. “I had to reopen the same tasks the next morning” explains why it matters. “Could we try a short update half an hour before the shift ends?” offers a change. Compare that with “Please communicate better,” which leaves both of you guessing about what to do next.
Make a fictional version before you open the chat
Use generic roles and change details that do not affect the practice. Omit names, internal messages, customer records, and allegations about identifiable people. In this scenario, you can say “a teammate sends information at the end of a shift” and still practise asking for an earlier update. The chatbot does not need access to your colleague’s actual messages or any team documents. Use an approved tool if workplace policy applies to the exercise.
Decide what the mock colleague can reasonably push back on. Perhaps they are still waiting for an update from another team. Perhaps the earlier time works only on some days. These are possible constraints to rehearse, not conclusions about why the person did anything. If the real situation involves a safety concern or conduct that requires a formal reporting channel, follow the applicable workplace process.
Run one exchange at a time
Try this original prompt with the fictional situation: “Help me rehearse a conversation about an end-of-shift handoff. Play a teammate who has a plausible reason for needing more time. Reply one turn at a time and let me respond. After three turns, tell me where my request was specific, where I assumed something, and what next step we agreed to. Point to my actual wording.”
Say your opening aloud if you can. If the first reply is “I can’t promise an earlier update because I’m waiting on the other team,” practise asking what can be shared before their answer arrives. A workable proposal could be an earlier note listing open items, followed by the missing detail later. For a second round, ask the chatbot to play a teammate who thinks the current arrangement works fine. Try acknowledging that perspective and then describing what happens on your side.
Use a simple scorecard for the next attempt
After the round, score your opening on three questions. Could a listener identify the event? Did you explain the impact without guessing at motives? Could the teammate repeat your proposed next step? Give each a simple yes or needs work. Ask the chatbot to point to the sentence behind any feedback. If it congratulates you without pointing to something specific, use the guide on how to tell whether an AI chatbot is just agreeing with you and give it a narrower question.
AHRQ’s TeamSTEPPS materials describe a structured way to raise a conflict in healthcare teamwork. The DESC script moves from describing the situation and expressing a concern to suggesting a change and spelling out its consequences. You can borrow that sequence as a rehearsal aid. In the handoff example the consequence is practical: more time to ask questions while both people are available. Treat the simulated response as a practice scenario. Use what you learn from your colleague in the actual conversation to adjust your approach.
Take one opening and one question into the room
Pick the version that sounds like something you would say. Keep an opening sentence and an honest question, such as “What usually has to happen before you can send the handoff?” Be ready to learn a constraint you did not rehearse. If you reach an agreement, restate who will do what and when you will check how it is working. If the original proposal is impractical, agree on the next step you can both try.
The same one-turn practice format can help in other contexts, including how to practise for a job interview with an AI coach. The aim here is narrower: show up with a clear request, give the other person room to explain what you missed, and leave with a next step that fits the real situation.
Sources
AHRQ TeamSTEPPS guide to the DESC script · NIST generative AI risk profile