To practice a sales call with AI, brief an AI buyer from the real account, run the call out loud, get scored on your team's scorecard, rerun the part that went wrong, then prepare the live call with a coach.
What the job is, who does it, how long it takes
Practicing a sales call with AI means rehearsing a specific upcoming conversation against an AI that plays the buyer, then getting feedback on what you said. It is the same idea as a roleplay with your manager, without waiting for the manager's calendar.
The rep does it, usually alone, in the day or two before a call that matters: a first discovery call with a senior buyer, a pricing conversation, a renewal where the champion has left. Managers use the same practice to get a new hire ready before they carry real accounts.
One practice loop takes 20 to 30 minutes. A five-minute warm-up roleplay, a short scored debrief, a second run on the weak part, then ten minutes preparing the real call. Spread across two days, it fits around a full calendar.
Why rehearsal works
Saying something out loud and getting it wrong in private is how most people get good at hard conversations. Learning research supports the instinct. In a classic study, people who practiced recalling material remembered 56% of it a week later, against 42% for people who re-read it (Roediger and Karpicke 2006). A roleplay is recall under pressure: the rep has to produce the question, in their own words, while the buyer pushes back.
Two other findings matter for how to practice. Spaced practice beats one long session, with an effect size of 0.46 across 63 studies (Donovan and Radosevich 1999), so two short rounds on two days help more than one long evening. And across 117 studies, training built on people's own scenarios, with trained managers involved, was linked to skills that held up (Taylor, Russ-Eft and Chan 2005). Practice on the real account, with the real objections, wherever you can.
Seven steps from brief to the real call
1. Pick the call and set one goal
Choose the next call that matters and decide the single thing you most need to get right. "Get the CFO to put a number on the cost of the problem" is a goal. "Have a good discovery call" is a wish. One goal per round keeps the feedback focused.
- 01Brief the buyerRole, priorities, likely objections
- 02Run it out loudBy voice, one goal per round
- 03Score itOn the team's own scorecard
- 04Rerun the weak partThe two minutes that went wrong
- 05Prep the live dealThe question to land, the trap coming
- 06Debrief afterWhat the buyer said, word for word
- 01Brief the buyerRole, priorities, likely objections
- 02Run it out loudBy voice, one goal per round
- 03Score itOn the team's own scorecard
- 04Rerun the weak partThe two minutes that went wrong
- 05Prep the live dealThe question to land, the trap coming
- 06Debrief afterWhat the buyer said, word for word
2. Brief the AI buyer from the real account
Give the AI buyer the facts you have: the buyer's role, the company's situation, what they said last time, the objection you expect. The closer the brief is to the real person, the more the practice transfers. Many teams keep personas for their common buyer types, so the rep only adds the deal-specific detail.
Do not make the buyer easy. A buyer who agrees with everything teaches the rep nothing. Ask for the skeptic: the CFO who has been sold to before, the operations director who thinks the current process is fine.
3. Run the call out loud, by voice
Speak the call. Typing a roleplay is useful when you cannot talk, but the real call is spoken, and the hard parts are spoken: the pause after a price, the follow-up question when the buyer gives a vague answer. Run the whole opening and the goal section, five to ten minutes, without stopping to restart.
4. Get scored on the team's own scorecard
Ask for feedback against the criteria the team already uses, whether that is a discovery scorecard, the stages of your method, or a short list your manager wrote. Feedback on the generic "build rapport, ask open questions" kind is easy to agree with and hard to act on.
A good score names the moment: the line where you answered your own question, the point where the buyer gave a number and you moved on without confirming it. Check that the scoring is consistent too. In our own research on AI call scoring, a redesigned scoring prompt moved up to 40% of scores when the same calls were scored again. A score that changes when nothing changed is noise. With our checklist model, only 1.7% of scores changed when the same calls were scored again.
5. Rerun only the part that went wrong
Skip the full replay. Go back to the two minutes that went wrong and run them again, two or three times, until the question comes out cleanly. This is where most of the improvement happens, and it is the step reps skip when they treat practice as a test to pass.
6. Prepare the live conversation with a coach
Practice teaches the delivery. The live deal needs a plan: which question opens the call, what the buyer has already said, what is still the rep's guess, where the trap is. Before the meeting, talk the deal through with a coach by voice or video on a call, five to ten minutes, and leave with the first question written down.
How to prepare a call with an AI coach covers that step in full.
7. Debrief after the real call
Within an hour of the real call, write down what the buyer said, word for word, on the goal you set. Compare it with the practice. If the buyer raised something the AI buyer never did, add it to the persona for next time. The debrief turns one call into better practice for the whole team.
A worked example
This example is illustrative. A rep at Arrowmere, a 600-person logistics firm, has a first meeting on Thursday with the finance director of a 300-person medical device maker. The finance director has agreed to 30 minutes and, according to the champion, thinks the current freight setup is "fine."
Wednesday morning. The rep sets the goal: get the finance director to put a number on what late deliveries cost. She briefs the AI buyer with the role, the "fine" comment, and the fact that the company switched carriers two years ago. She asks for a skeptic.
First run. She opens with a slide's worth of Arrowmere's capabilities. The AI buyer says the company already has a carrier and asks why this is a meeting. She recovers, asks about late deliveries, and gets a vague answer: "Some weeks are worse than others." She moves on.
Score. The feedback names two moments. She led with Arrowmere before the buyer had said anything about a problem. And when the buyer said some weeks were worse, she did not ask what a bad week costs.
Rerun. She runs the opening three times. By the third, she opens with what the champion told her about missed hospital delivery windows and asks how the finance team accounts for them. When the AI buyer says "some weeks are worse," she asks, "What did the worst week last quarter cost you?"
Wednesday afternoon. She talks the live deal through with Caty, our AI sales expert for any method, in a short call. Caty asks what the champion has done for the deal so far and whether the finance director knows the champion asked for the meeting. The rep writes down her first question and a fallback.
After the meeting. The finance director gives a number for missed delivery windows. He also raised a contract renewal date the AI buyer never mentioned, so the rep adds it to the persona.
Where practice stops
Practice gets the words right. It cannot tell the rep whether the deal is real, who signs, or what the buyer said on the last call that changes the plan. Those questions belong to the live deal, and they need someone who knows the team's method and the account.
That is the line between roleplay and deal coaching, and AI coaching vs AI roleplay sets the two side by side. Pure roleplay platforms are good at the practice half, and our comparison with Hyperbound describes one fairly. Caty does both in one place: roleplay against a buyer built from your personas, scored on your criteria, and a call on the live deal before the meeting and after it. The AI sales coaching guide covers choosing between tools.
Where it goes wrong
- The easy buyer. An AI buyer who agrees with everything gives the rep false confidence. Brief a skeptic every time.
- Typing the whole thing. Typed practice skips the hardest part of a real call, which is saying the question out loud under pressure. Type only when you cannot talk.
- Generic scoring. Feedback that could apply to any call changes nothing. Score against the team's own criteria and ask for the moment.
- Replaying the whole call. Running the full call again and again rehearses the parts that already work. Rerun the weak two minutes.
- Stopping at practice. A perfect roleplay on a deal with no real buyer pain is still a deal with no real buyer pain. Prep the live deal too.
To set this up for a team, talk to us.
FAQ
Can AI help you practice sales calls?
AI can play the buyer in a realistic roleplay, push back with objections from your real accounts, and score each attempt against your scorecard, so reps can rehearse without waiting for a manager.
How long should a sales call practice session be?
A practice loop of 20 to 30 minutes is enough for most calls: a five to ten minute roleplay, a scored debrief, a rerun of the weak part and ten minutes preparing the live deal.
Should I practice sales calls by voice or by typing?
Practice by voice whenever you can, because the real call is spoken and the hard moments are spoken. Typing is the fallback for when you cannot talk, such as on a train.
What should an AI buyer persona include?
An AI buyer persona should include the buyer's role, the company's situation, what they said last time, their priorities and the objection you expect. A skeptical buyer teaches more than an agreeable one.
Is AI roleplay enough to prepare for a real sales call?
Roleplay prepares the delivery, and the live deal needs a plan as well. Before the call, decide the first question, what the buyer has already said, and what is still your guess.