How to run cold call roleplay that reps do not hate

Ask a room of SDRs what they think of roleplay and you will get a groan. Ask their managers and you will get a shrug: they know it works, they know nobody does it, and they have stopped scheduling it. Both are right. Roleplay as it is usually run is a bad exercise, and the fix is not to do more of it but to change what it is.

Why the usual version fails

There is an audience. The standard format is one rep, one manager playing the prospect, and the rest of the team watching. The rep is not practising a cold call; they are performing one, in front of the people whose opinion matters most to them. Nobody takes a risk in that setting. They run the safest version of their script and hope it ends quickly.

The manager plays the prospect. Managers make bad prospects. They know the product too well, so they raise the objections the pitch is designed to handle. They know the rep, so they either go easy or, worse, decide today is the day to make a point. And because they are playing the prospect, they cannot also watch the rep. They come out with an impression rather than notes.

There is no repetition. A rep gets one turn, maybe two, in a session that happens every other week if the calendar holds. Nobody has ever got good at anything with two attempts a fortnight. Compare it with the real thing: a rep might make fifty dials a day. Roleplay at that cadence is not practice, it is assessment.

Nothing is at stake, and everything is. The prospect always caves eventually, because the manager wants to get through the exercise. So the rep learns that persistence gets rewarded. Meanwhile the social stakes are high, so the rep learns to dread it. The incentives are exactly backwards.

What to change

Take the audience away. Practice should be private by default. A rep should be able to run the same objection twelve times in a row, badly, without anyone watching. Share the good one afterwards if it is useful to the team, but the repetitions and the failures stay between the rep and whoever is coaching them.

One skill per session. Do not run "a cold call". Run the first thirty seconds, twenty times. Run "we already have a solution", ten times. Run the close, from the point where the prospect has softened. A full call is too much to change at once, and the feedback ends up being a list of eight things, of which the rep will remember none. Pick the dimension with the lowest score and drill that until it moves.

Use levels. A new rep should start against a prospect who is friendly and has time, because the point is to get the words out in order. Once the words are automatic, the prospect gets busier and more skeptical. Only then do you introduce the senior buyer who was interrupted mid-task and wants you gone. Jumping straight to the hard prospect teaches new reps to freeze; staying on the easy one teaches tenured reps nothing.

The prospect decides. The most important change. In real life the rep does not get to decide whether the call worked; the prospect does, and they decide based on what was actually said. Roleplay should work the same way. The prospect should agree to a next step only if it was earned, and the session should end with a clear outcome: meeting, callback, or no. A roleplay where the prospect always comes round is a rehearsal for a world that does not exist.

Scoring it consistently

Roleplay feedback is usually a mix of "that was good" and whatever the manager happened to notice. Two managers watching the same call will give different notes. The same manager will give different notes on Monday and Friday. That inconsistency is why reps do not trust it.

Score every call against the same rubric, and make it short. Six dimensions is about right: opener, reason for the call, discovery, objection handling, value, close. For each, a number and one sentence that quotes what the rep said. "Objection handling, 4: when the prospect said 'we've got something', you went straight to features instead of asking what they had."

Two rules make the scores mean something. Score from the transcript, not from memory; if you cannot point at the line, it does not go in the notes. And be stingy. If a nine is common, the score tells the rep nothing. The scale should leave room above where your best rep sits today.

Then track the dimension over time, not the call. Five calls with the same weak dimension is a coaching priority. One bad call is a Tuesday.

Where AI practice changes the maths

Everything above can be done with a manager, a rep and a rubric, and it should be. What it cannot be is done at volume. A manager with eight reps who runs one twenty-minute session per rep per week is spending nearly three hours to give each rep one attempt.

An AI prospect changes the arithmetic without changing the method. The rep can run the same objection fifteen times before lunch. The prospect is built from real target accounts, so it raises the objections those buyers raise and not the ones in the pitch deck. It does not know the rep, it does not go easy, and it decides the outcome from what was said. Every call comes back as a transcript with the prospect's decision, and the rep can ask for a scored coach's review when they want one.

What it does not do is replace the manager. The rep still needs someone to read the transcript with them and say "here, this line, try it this way". What changes is that the manager arrives at that conversation already knowing which dimension is weak, which objection keeps going wrong, and the exact line where it happens. The twenty minutes are spent coaching rather than listening to find something to coach. If you are building a practice programme for a team, the enablement page describes how that fits together.

A session that works

Fifteen minutes. The rep picks one dimension based on last week's scores. They run five calls against a prospect one level harder than they are comfortable with, with the goal of moving that one score. Then they read the transcript of the worst of the five and write down the line they would change. Once a week, they bring that line to their manager.

No audience, one skill, real resistance, a consistent score, and enough repetitions to learn from. Reps do not hate that. They hate being watched, judged inconsistently and given one turn.