#60 Blog. The Best Prompt for Manager Simulations. Why the way you set up your AI role-play determines what you actually learn from it.
Last week, I was testing a coaching scenario with AI. You probably tried it already yourself.
I typed something like: "Simulate a difficult conversation with an underperforming employee. This is the 3rd time you are about to talk with him about it."
And the AI played along.
Too well, actually.
The "employee" apologized immediately.
Agreed with everything I said.
Thanked me for my honesty.
The conversation ended in about 45 seconds, and I felt... good actually.
Confident. Like I had just nailed a mid-year review.
I hadn't nailed anything.
I had written a lazy prompt, and I got a lazy simulation back.
And I think you should know by now: the best simulation doesn't start with the AI. It starts with your prompt.
Not the first time you hear about the importance of prompt right?
If you've read my last few posts, you know I'm a big believer in practice, not just learning. But practice is only as good as the scenario you practice in. A soft, easy, agreeable simulation doesn't build your muscle. It just makes you feel good for five minutes, and then what? Do you feel equipped for the real-life scenario?
So today, let's talk about something a little more practical: how do you actually write a prompt that gives you a real, useful, uncomfortable-in-the-right-way simulation?
Why most prompts fail before the conversation even starts
Most managers, or learning and development professionals, approach AI simulations the same way they approach a real conversation they're dreading: vaguely, and hoping it just works out.
"Pretend to be a difficult employee."
"Simulate a performance conversation."
"Act like my manager and give me feedback."
These prompts aren't wrong. They're just empty. There's no tension in them. No personality. No stakes. And a simulation without tension can't teach you anything, because the real conversation you're preparing for is never that clean.
You don't need a perfect prompt. But you do need a specific one.
Bad Prompt vs. Good Prompt
Bad Prompt
"Simulate a hard conversation with my employee."
Generic role, no backstory
No personality, no resistance
Ends the moment you say the "right" thing
Feels good. Teaches you nothing.
Good Prompt
A specific person, in a specific moment, with a specific history
A personality that pushes back
Real stakes for both of you
Doesn't fold just because you said the right words
Feels uncomfortable. Teaches you something.
That's the difference. A bad prompt gives you a performance. A good prompt gives you a rehearsal.
So what you must have in a good prompt?
Think back to that example I gave you a few blogs ago- the employee who's vocal, interrupts people, and works in customer service, and you need to raise it with him. Let's actually build that simulation properly this time.
1. Give the AI a role, not a label.
"Difficult employee" tells the AI almost nothing. Instead, try:
"You are Daniel, 34, a customer service rep on my team for two years. You're talented, well-liked, but you interrupt people constantly, including customers. You don't think it's a big deal. You get a little defensive when it's brought up, but you're not hostile- you genuinely don't see the problem yet."
Now the AI has someone to be.
2. Tell it what "success" and "failure" look like, from the character's side.
AI needs direction just like people do.
Add: "You should only start to soften and actually reflect if I ask questions instead of accusing you, and if I acknowledge what you're good at before raising the issue. If I come in too harsh or vague, stay defensive and change the subject."
This one instruction alone will save you from that 45-second "nailed it" conversation you and I had so far.
3. Set the emotional temperature.
Is this the fifth time you're having this conversation, or the first? Is trust high or low between you two? Tell the AI. A prompt like "this is the third time this has come up, and you're starting to feel like I'm nitpicking you" changes everything about how the character responds.
4. Ask for feedback afterward. Every time.
The simulation isn't actually the valuable part. What you learn from it is. End every practice session with:
"Now step out of character. As my coach, what worked in how I opened this conversation? What would you have done differently? What did you notice about my tone?"
This is the step most managers skip, and it's the one that turns a fun exercise into actual growth. Where the real learning happen.
5. If it wasn't clear- use the audio option, and stop writing. 😊
A quick note on discomfort
If your simulation ends easily, you probably wrote a prompt that was too soft. And I get it- we don't build these scenarios to torture ourselves. But remember: would you rather feel a little uncomfortable now, with an AI, or feel that same discomfort for real, with someone whose trust you actually need?
The AI doesn't remember the conversation tomorrow. Your employee will.
You wouldn't walk into a real performance conversation without thinking about the person across the table- their history, their personality, what might make them defensive, what might make them open up.
Don't walk into your practice conversation without doing the same.
The quality of your simulation is a direct reflection of the quality of your prompt. Vague prompt, vague growth. Specific, honest, slightly uncomfortable prompt- real growth.
Here's the thing- I didn't learn this by accident.
When we built RiseBud, this was exactly the problem we had to solve.
Every scenario inside RiseBud is grounded in proven leadership frameworks to build the psychological safety- not just "pretend to be difficult." And because the conversation happens live, the AI is genuinely listening and responding to what you say, in real time. So even if you run the exact same scenario twice, you'll never get the exact same conversation. Say something clumsy, and it reacts like a real person would. Open with real curiosity, and he/she opens up differently too. There's no script running underneath it- just like there isn't one in your real 1:1. Just a really good prompt ;)
And when the conversation ends, you don't just walk away with a feeling. You get detailed feedback on what actually happened- what worked, what didn't, what to try differently next time. Which, if you remember step 4 above, is exactly the habit that turns practice into growth. We just built it in, so you don't have to remember to ask.
And it doesn't forget you the moment you close the tab. RiseBud remembers your past conversations- so if you keep struggling with the same kind of feedback, or the same kind of employee, it knows. Next time you open a scenario, you're not starting from zero. It's building on what it already knows about how you communicate, where you tend to get stuck, and what you've already worked through. That's the difference between a simulator you use once and a coach you actually grow with.
I think the real game-changer, though, is that you get to choose the level of difficulty- easy, medium, hard. What's the level of conversation you actually need to have this week? Remember- nobody else is listening 😊
So no, you don't have to become a prompt engineer to get a realistic, adaptive, ever-changing practice conversation with real feedback at the end. If you're experimenting with your own scenario in another tool, the five elements above will get you started.
If you need some supprt- try RiseBud.
Need help building your next scenario- let's talk.
P.S. If this one gave you an idea for a conversation you've been avoiding, I'd love to help you build the prompt for it. Just reach out.
Hana
Comments