You've been putting off a hard conversation for a week. Every time you think about it, you either freeze or rehearse an angry version in your head that won't actually help.
An AI chat tool can give you a safer place to rehearse it first — one where you can try it out loud, get it wrong, and try again before it matters.
Quick summary
- AI chat tools can role-play the other person in a hard conversation.
- Rehearsing out loud, even with an AI, changes how a conversation actually goes.
- Ask for a calm version and a firm version, then pick what fits you.
- The AI's replies are guesses, not proof of how the real person will react.
Why this matters
Hard conversations feel worse in your head than they usually are in real life, because you're imagining every possible reaction at once. Rehearsing narrows that down to something you can actually practice.
Saying words out loud, even to an AI, is different from just thinking them. It catches phrasing that sounds harsh or unclear before you use it on a real person.
How to do it
- Set the scene first. Tell the AI who the other person is, what happened, and what you want from the conversation.
- Ask it to play the other person, realistically. "Respond the way you think they might, including some pushback" gets more useful practice than a purely agreeable reply.
- Ask for feedback on your opening line specifically. The first sentence sets the tone for the whole conversation, so it's worth getting right.
Run through it two or three times. Notice which version feels most like something you could actually say, not just something that reads well on the screen.
A real example
Say you need to tell a coworker their comments in meetings come across as dismissive. You could open with: "I want to run through a conversation before I have it. Play the role of my coworker Alex, who tends to get defensive. I'm going to tell them their comments in meetings sometimes shut other people down." The AI's reply gives you a rough version of the pushback you might face — so your real opening line isn't the first time you've heard it.
Try a calm version and a firm version
Most hard conversations have more than one right tone, depending on how the other person tends to respond. It helps to rehearse two versions rather than settling on the first one that comes out.
Ask the AI to give you a calm, collaborative opening first — something that assumes good intent and leaves room for the other person to explain themselves. Then ask for a firmer version, one that states the issue plainly with less cushioning. Read both out loud. You'll usually notice one fits how you actually talk, and the other sounds like someone else's script.
Where this approach can go wrong
Rehearsing too many times can backfire. After the fourth or fifth run-through, some people start sounding scripted and stiff instead of natural, because they're reciting rather than speaking. Two or three passes is usually the useful range — enough to find calmer wording, not so many that it stops sounding like you.
It's also worth remembering the AI doesn't know the real person. It's guessing at a plausible reaction based on what you've described, not predicting the actual conversation. Use it to sharpen your own delivery, not to convince yourself you already know how they'll respond.
Rehearsing conversations that involve more than one person
Not every hard conversation is one-on-one. Team disagreements, family gatherings, and group decisions often involve several voices at once, and it's harder to picture how each person will react. You can still use AI for this — just describe each person separately, with their likely stance, before asking the AI to play the group.
Ask it to respond as the most resistant person first. That's usually the reaction you're most anxious about, and getting a rough version of it out loud takes some of the edge off. Then ask for a second pass where the AI plays someone more neutral, so you can hear how the same opening line lands differently depending on who's listening.
What this method won't do for you
Rehearsing sharpens delivery, but it doesn't replace listening once the real conversation starts. A common mistake is walking in with a rehearsed script and sticking to it no matter what the other person actually says, because it feels safer than improvising. Treat your rehearsed opening as a starting point, not a script to defend. The goal is a calmer first thirty seconds, not a performance you have to finish word for word.
It also won't fix a conversation that needed to happen weeks ago. Rehearsal reduces anxiety about how to say something — it doesn't reduce how overdue that something has become. If you've been avoiding this for a long time, the AI practice helps, but picking a specific day to actually have the conversation matters more than any extra rehearsal.
Try this now
Pick one conversation you've been avoiding. Describe the situation to an AI chat tool and ask it to play the other person. Run it twice, then write down the opening line that felt most natural.
Frequently asked questions
Will the AI predict exactly how the real person reacts?
No. Treat it as practice for your own delivery, not a forecast of their actual response.
Is this only useful for work conversations?
No — it works for family, friends, or any conversation where you want to find calmer wording first.
What if I don't know what I want to say yet?
Start by just describing the situation and asking the AI what the core issue seems to be. That often clarifies what you actually want before you rehearse it.