How to practise sales calls with AI: prompts and a scoring rubric

You have a renewal meeting with a finance director on Thursday and nobody to practise with. Your manager is busy, your colleagues sell different products, and rehearsing in the car only gets you so far. A general AI chatbot can play the buyer, if you set it up properly.

Below: the prompts, a scoring rubric built on the seven principles of The Pharaoh's Pitch, and an honest account of where a general chatbot stops being useful.

Can you really practise sales calls with a chatbot?

Yes, for the words. A chatbot is a good sparring partner for the questions a buyer asks, the objections they raise and the way you phrase your answers. It is weak on everything that isn't words: tone, pauses, interruptions and the look on someone's face when you mention price.

That still covers a lot. Deals tend to turn on a few moments you can see coming: the price question, the "we're happy with our current provider" brush-off, the request to just send something over. Moments you can predict are moments you can practise.

The book behind this blog describes the same habit without the technology. In chapter 5, Ivan Yong lists mental rehearsal among the self-leadership strategies a salesperson can use, and describes how he does it before client meetings:

I create a mental picture of potential questions that my clients will ask and how they will ask these questions, including the mood they are in.

An AI buyer is that mental picture, made external. It asks the questions out loud, in the mood you gave it, and it doesn't let you skip the hard ones.

Step 1: Brief the AI buyer properly

The quality of the practice depends almost entirely on the brief. A one-line prompt ("pretend to be a tough CFO") tends to produce a generic, oddly polite buyer who folds quickly. A good brief gives the AI a role, a situation, real concerns and a personal motive it keeps to itself.

That hidden motive is the part most people leave out. The book introduces the idea of a "personal win" in chapter 2: something a buyer wants for themselves from the purchase, which may have little to do with what their company wants. In the book's examples it is things like recognition in their field or a move to a bigger employer.

Give the AI a personal win and tell it not to volunteer it, and you are practising the hardest part of discovery.

The buyer brief (copy, fill in, paste as your first message)

"You are going to play a buyer in a sales role-play. Stay in character until I type END. Who you are: [role] at [type of company, size, location]. Your decision style: [for example, cautious and wants detail, or quick and wants the bottom line]. The situation: [what I'm selling, what stage the deal is at, what happened last time we spoke]. Your concerns: [two or three real concerns]. Your personal win: [what you want for yourself from this decision]. Do not reveal it unless my questions earn it. Rules: reply in two to four sentences, like a busy person. Push back on vague answers. Don't agree to anything just because I'm polite. If I handle a concern well, move to the next one; if I handle it badly, press harder. Don't coach me or comment on my performance until I type END."

Here is the middle of that brief filled in for an insurance broker:

"Who you are: managing director of a family-owned logistics company in Shah Alam, about 80 staff. Practical, suspicious of salespeople. The situation: I'm a broker proposing to move your group medical cover to a new insurer at renewal. Your concerns: premiums went up last year, staff complained about slow claims, and switching means paperwork your HR person has no time for. Your personal win: your daughter joins the business next year and you want to hand her a company that runs properly."

For a SaaS seller in Singapore, swap in a finance director at a regional distributor whose personal win is getting through the next audit without the month-end scramble.

Step 2: Run the call like a real one, then run it again

Treat the role-play as a call, not a chat. Type what you would say out loud, at the length you would say it. A three-paragraph reply to "why are you more expensive?" is practice at writing, not selling.

A few ways to get more out of each session:

• Practise one moment, not the whole call. Open with: "Start the conversation at the point where I've just shared the price."
• Don't restart when it goes badly. Recovering from a weak answer is the skill.
• Change the mood and go again. "Same scenario, but you're short on time and irritated because your CFO has just frozen new spending."
• Change the buyer, keep the objection. A procurement manager and a founder push back on price in very different ways.

That last point is where the book adds a caution. Chapter 5 recommends self-rehearsal for people starting out, with a warning attached:

Self-rehearsal could be an important strategy if you are new to sales, but I caution against over-rehearsing.

His reason: too much rehearsal can upset the flow of a real conversation, and it is better to stay nimble. So don't memorise one perfect answer. Use it to hear the same objection five different ways, until you can handle whichever version turns up.

Step 3: Score the call with a rubric built on the seven principles

A rubric is how you stop the AI from telling you that you did brilliantly. When you type END, paste this:

"Step out of character. Score my performance from 1 to 5 on each criterion below. For every score, quote my exact words as evidence. If you can't find evidence, score it 1. Then show me the one moment where I lost the most ground and what I could have said instead. Don't soften the feedback."

Then paste the criteria. Each one comes from a principle in the book:

The evidence rule matters more than the numbers. Chapter 1 of the book warns about a related trap with personality reports: once people read one, they tend to agree with all of it, even the parts that would fit anyone.

AI feedback has the same pull. "You built good rapport" feels true and tells you nothing; a quoted line you can reread does.

If you'd rather not assemble this yourself, Niles has a role-play tool: the AI plays the client, then debriefs you, drawing on the same book as this rubric.

Where a general chatbot falls short

A general chatbot is a decent rehearsal partner and a poor judge. Keep these limits in mind:

• It is polite by default. Left alone, the AI buyer warms up, accepts a so-so answer and moves on. The rules in the brief help, but you will still need to nudge it: "You wouldn't accept that answer. Push back."
• It doesn't know your deal. It knows only what you typed about the buyer, your product and your market. A thin brief produces generic objections, and you will often find yourself pasting the same context again each session.
• It can't hear or see you. Chapter 7 points out that people rarely say how they feel, so empathy depends on reading non-verbal cues. A typed conversation gives you none, and nothing tells you that you rushed, mumbled or went quiet after the price.
• It has no ground truth. It can say an answer was weak, but not whether it would have worked on this particular buyer. Without a fixed rubric, its idea of a good answer can also drift from one session to the next.
• It can sound sure and be wrong. If the AI buyer raises a regulatory point or a claim about a competitor, check it before you repeat it to a client.

None of this makes the practice worthless. It means AI role-play is for rehearsing your words and your thinking, and the real call is still where you find out. If you're weighing a tool against a person for the rest, our comparison of an AI vs human sales coach covers what each can and can't do.

After the real call: debrief, then practise what actually came up

Practice is half the loop. The book's epilogue asks you to debrief yourself straight after every sales meeting, including the ones that went against you: write down what happened objectively, without emotion, then reread chapter 6 and let it point you to the chapter that explains what went wrong. Chapter 5 describes a similar habit, self-monitoring: reviewing your own behaviour and the client's reactions after a meeting, then deciding what to change.

AI makes that faster. Within an hour of the call, paste in your notes with this prompt:

"Here are my notes from a real sales meeting: [notes]. Using the same rubric as before, tell me where I'm probably marking myself too kindly, what the buyer's real concern seems to be, and which objection I should practise before we meet again."

Then take the objection the buyer actually raised and run it through Step 2: the real buyer supplies the material, and the chatbot supplies the repetitions. Combine it with how to prepare for a first sales meeting and you have a full prepare, practise and debrief cycle.