10 AI Roleplay Prompts for Sales Teams in 2027, Built From Real Deals

A good AI roleplay prompt for sales gives the AI buyer a role, a hidden agenda, rules for what it reveals, and a rubric for scoring the rep, all taken from your real deals. The ten copy-ready prompts below cover discovery, status quo, CFO pushback, a competitor's customer, the economic buyer, champions, committees, security reviews and procurement, plus a debrief prompt that scores the rep.

Key takeaways

• Realistic roleplay comes from real material: the objection wording, competitor and stakeholder mix from your last 30 days of calls.

• Every prompt needs reveal rules. A buyer who volunteers everything teaches reps to stop asking questions.

• Score every session against a short rubric, then rerun the same scenario until the score moves.

• Rotate scenarios monthly as new objections and competitors appear in your pipeline.

• In Proshort, Aria scores skills from real calls, builds role-play drills for the gaps, and reps practice against an AI buyer that pushes back like real buyers.

Why do sales teams need better AI roleplay prompts?

Reps want practice and are not getting it. In Salesforce's State of Sales report (7th edition), over half of reps say courses and guides do not give them the skills they need, and many say they get too little feedback on calls or chances to practice important conversations ahead of time. The same report says about a third of sales teams using AI agents use them for coaching, for example to roleplay sales calls.

The weak link is usually the prompt. "Pretend to be a tough buyer" produces a polite buyer who folds after one answer. A rep who practices against that buyer learns the wrong lesson.

What makes an AI roleplay prompt realistic?

Anthropic's prompting guide notes that setting a role focuses the model's behavior and tone, and that XML tags help it separate instructions from context. OpenAI's prompt engineering guide suggests structuring instructions as identity, instructions, examples and context. For roleplay, that becomes six parts:

1. Role: title, company, industry and what the buyer cares about personally.

2. Hidden agenda: the real pain, budget or politics the rep must uncover.

3. Reveal rules: what the buyer shares only when asked a specific, open question.

4. Behavior: short answers, interruptions, skepticism, time pressure.

5. End condition: when the call ends (for example, after the rep proposes a next step).

6. Scoring: three to five criteria the debrief prompt will grade.

Each prompt below ends with the same guardrail line: "Stay in character. Answer in one to three sentences. Do not coach me during the call." Paste it into ChatGPT, Claude or your roleplay tool as the system or first message, then start the call yourself.

The 10 roleplay prompts at a glance

Ten AI roleplay prompts for sales reps

1. The evasive operator (discovery)

When to use it: New-hire onboarding, or when deals reach demo without a clear pain.

You are [title] at a [industry] company with [size] employees. Your real problem is [pain from a real call], which costs you [impact], but you will not say so unless I ask a specific follow-up question about it. Answer surface questions with vague, true-sounding answers. End the call when I propose a next step. Stay in character. Answer in one to three sentences. Do not coach me during the call.

What good output looks like: The buyer gives nothing to "What keeps you up at night?" and opens up only after the rep asks about a specific process, a number or a consequence.

Common mistake: Writing the pain into the opening line. If the buyer states the problem, the rep practices pitching, not discovery.

2. The status quo defender

When to use it: Early-stage deals where the real competitor is "do nothing".

You are [title] at [company]. You recently invested in [current process or tool] and believe it works well enough. You see no reason to change this year. Push back on every claim of value with "we are fine as we are" until I quantify what the current way costs you, using numbers you give me only when asked. Stay in character. Answer in one to three sentences. Do not coach me during the call.

What good output looks like: The rep asks for the buyer's own numbers and builds the cost of inaction from them, instead of quoting generic benchmarks.

Common mistake: Letting the buyer accept urgency after one statistic. Real status-quo buyers need their own numbers.

3. The skeptical CFO

When to use it: Before any business case, pricing or executive meeting.

You are the CFO of [company]. You have 15 minutes. Challenge every ROI claim: ask for the assumptions, the payback period, what happens if adoption is half of plan, and what we will stop spending to fund this. Reject vague answers such as "significant savings". Stay in character. Answer in one to three sentences. Do not coach me during the call.

What good output looks like: The rep states assumptions, shows a conservative case and asks what return the CFO needs, rather than repeating the deck.

Common mistake: Practicing with numbers the rep cannot defend. Use the same ROI model the rep will show the real buyer.

4. The competitor's happy customer

When to use it: Rip-and-replace deals and competitive bake-offs.

You are [title] at [company] and have used [Competitor] for two years. You like it. Raise the switching costs: migration effort, retraining, contract timing and the risk of a worse result. Only consider switching if I find a gap that matters to you, which is [real gap from won or lost deals]. Stay in character. Answer in one to three sentences. Do not coach me during the call.

What good output looks like: The rep asks what the buyer wishes were better before pitching, and talks about migration risk plainly.

Common mistake: Making the competitor look weak. Reps then freeze when a real buyer defends the incumbent.

5. The elusive economic buyer

When to use it: Mid-stage deals where MEDDICC has gaps in economic buyer, decision criteria or decision process.

You are the economic buyer at [company]. You give short, noncommittal answers and redirect budget questions to "the committee". Reveal your metrics, decision criteria, decision process and timeline only if I ask a specific open question about each. Keep track of which ones I never asked about. Stay in character. Answer in one to three sentences. Do not coach me during the call.

What good output looks like: At the end, the buyer lists which MEDDICC elements the rep confirmed and which were never asked.

Common mistake: Running qualification as a checklist. The skill is earning each answer, not reading the acronym aloud.

6. The champion without a business case

When to use it: Deals with one enthusiastic contact and no executive access.

You are a [title] who likes our product but must present it to [executive] next week, and you do not know how to build the case. Ask me for the three points to lead with, proof from similar companies, and how to answer [executive]'s likely objection. Push back if my advice would not survive your internal meeting. Stay in character. Answer in one to three sentences. Do not coach me during the call.

What good output looks like: The rep coaches the champion: a short story, one proof point, the likely objection and an offer to join the meeting.

Common mistake: Treating the champion as the decision maker. The drill is about arming them, not closing them.

7. The conflicted buying committee

When to use it: Enterprise evaluations with three or more stakeholders.

Play three people on one call: [IT lead] cares about security and integration, [finance lead] cares about cost, and [operations lead] wants ease of use and is enthusiastic. Respond as whichever person my question affects, and let them disagree with each other. End when I propose a next step that addresses all three. Stay in character. Answer in one to three sentences. Do not coach me during the call.

What good output looks like: The rep names each stakeholder's priority out loud, assigns follow-ups per person and does not let the friendliest voice dominate.

Common mistake: Using a committee that agrees. Real committees argue, and the rep has to manage that.

8. The persistent security reviewer

When to use it: Before security questionnaires, IT reviews or regulated-industry calls.

You are the [CISO or IT security lead] at [company] in [industry]. Each time I answer, raise a new related concern: data residency, access controls, retention, subprocessors, then exit terms. Ask for documentation by name. If I guess, say so. Stay in character. Answer in one to three sentences. Do not coach me during the call.

What good output looks like: The rep answers what they know, admits what they do not, and commits to a dated follow-up with the right expert.

Common mistake: Rewarding confident guesses. In security reviews, a wrong answer costs more than "I will confirm by Thursday".

9. The last-minute procurement curveball

When to use it: Late-stage and quarter-end deals.

You are a procurement manager at [company]. The contract is ready to sign. Ask for a [percent] discount "to get it approved this quarter", then add new payment terms. Halfway through, mention that a competitor offered a lower price. Concede only if I trade for something of value. Stay in character. Answer in one to three sentences. Do not coach me during the call.

What good output looks like: The rep asks what the discount is for, trades instead of giving (term length, a case study, a faster signature) and protects price.

Common mistake: Practicing only the discount. The second ask after the first concession is what real procurement does.

10. The debrief coach

When to use it: After every roleplay session, in the same chat.

Step out of character. Score my performance from 1 to 5 on [criteria, for example: discovery depth, handling the main objection, confirming next steps]. For each score, quote the moment in our conversation that justifies it, then give me the one better line I should have said. Finish with the single scenario I should replay next.

What good output looks like: Scores tied to quoted moments, one better line per criterion and a clear next drill.

Common mistake: Asking "how did I do?". Without criteria and quotes, the feedback is generic praise.

How do you build roleplay scenarios from real pipeline and CRM data?

Fill the placeholders from your own deals, not your imagination. Pull the three most common objections from last month's calls in the buyer's exact words, the competitor named most often in open deals, and the loss reasons from the last quarter. Use the stakeholder mix from a real late-stage deal for prompt 7. Rotate the set monthly, and raise difficulty as scores improve.

Measure the program with a few numbers you already track: ramp time for new hires, win rate in deals where the practiced objection appeared, and stage conversion for the reps who practiced. If those do not move within a quarter, the scenarios are too easy or too far from what buyers actually say.

How does AI role-play work in Proshort?

Proshort's skills and AI role-play scores reps on skills such as MEDDICC, negotiation and stakeholder mapping from their real calls, builds a plan with role-play drills, and lets reps practice against an AI buyer that pushes back like real buyers. Coaching comes from the calls themselves: each call is scored against your own playbook, and the rep sees the moment, the timestamp and the better line, then rehearses it with Aria.

Managers can ask Ask Aria "Who needs coaching this week?" and schedule the answer as a daily 8:00 email. And when the real call comes, real-time objection handling detects the objection live and shows the rep the answer that has won before, privately, during the call.

"It's effectively giving them on demand coaching without having to seek out Brian or myself to help them out with that." Elias Reynoso, Head of Revenue Operations, Vitable Health (read the Vitable Health case study)

Comparing platforms rather than prompts? See our roundup of AI roleplay tools and the best AI sales training software.

Frequently asked questions

What is an AI roleplay prompt for sales?

It is a set of instructions that turns an AI model into a realistic buyer for practice. A good one defines the buyer's role, hidden agenda, reveal rules, behavior, end condition and scoring criteria, so the rep has to earn information the way they would on a real call.

Can I run these roleplay prompts in ChatGPT or Claude?

Yes. Paste the prompt as the first message, fill in the placeholders, and type or speak your side of the call. Use prompt 10 at the end of the same chat to get a scored debrief with quotes from the conversation.

How do you make AI roleplay scenarios realistic?

Use material from your own pipeline: objections in the buyer's exact words, the competitor you meet most, and stakeholder mixes from real deals. Add reveal rules and short-answer behavior so the AI buyer does not volunteer everything.

How often should reps do AI roleplay?

A short session before any high-stakes call works better than a long monthly workshop. Many teams add one or two scenarios a week for new hires and one targeted drill a week for tenured reps, based on gaps found in real calls.

How do you measure whether AI roleplay is working?

Track ramp time for new hires, win rate in deals where the practiced objection appeared, and stage conversion for reps who practiced. Roleplay scores should rise first; deal metrics should follow within a quarter.

See how reps drill the gaps found on their own calls against an AI buyer: book a Proshort demo.