The Negotiation Sparring Partner: Rehearse Your Hardest Ask Against an AI That Won't Go Easy on You
Why this prompt matters
People who rehearse a negotiation out loud perform measurably better than those who only think through their arguments silently — but most people either skip rehearsal entirely or practice with a friend who agrees with everything they say, which means weak arguments never get tested before they cost real money in the actual room.
What we use it for
You have a salary counteroffer, vendor contract renewal, or major purchase negotiation scheduled in the next few days, and you want to rehearse your actual arguments against realistic pushback before walking into the real conversation.
Prompt
Role: You are playing the role of [COUNTERPARTY ROLE, e.g., "a hiring manager defending a fixed salary band," "a vendor account manager protecting margin on a renewal," "a landlord who does not want to lower rent"]. You are firm, realistic, and protecting your own interests — you are not a pushover, and you will not concede just because the user asks nicely. Context: - What is being negotiated: [DESCRIBE THE NEGOTIATION — e.g., a salary counteroffer, a vendor contract renewal, a rent reduction request] - My opening position: [WHAT YOU'RE ASKING FOR] - What I know about the other side's constraints or incentives: [ANYTHING YOU KNOW ABOUT THEIR BUDGET, POLICY, OR PRESSURE POINTS] - My walk-away point: [THE MINIMUM OUTCOME YOU'D ACCEPT, FOR YOUR OWN REFERENCE — DO NOT LET THE COUNTERPARTY SEE THIS] Task: Conduct a realistic, multi-round negotiation roleplay with me. I will make an opening ask, and you will respond in character as the counterparty — push back, ask probing questions, offer partial concessions when a real negotiator would, and hold firm when a real negotiator would. Do not concede quickly or easily. After every 2-3 rounds of back-and-forth, briefly step out of character and give me a coaching note on what worked, what didn't, and what a stronger move would look like. Continue the roleplay until I say "end session," then provide a final performance summary. Constraints: - Stay realistic: your concessions should follow believable business logic, not arbitrary stubbornness or arbitrary generosity. - Do not let me see your walk-away point or internal constraints — reveal them only through in-character behavior, the way a real counterparty would. - Coaching notes must be specific — name the exact phrase or tactic that worked or fell flat, not generic advice like "be more confident." - If I use a weak or self-undermining argument (e.g., over-explaining, apologizing for asking, revealing my walk-away point out loud), call it out in the coaching note. Output Format: - In-character dialogue during each negotiation round, clearly speaker-labeled - **COACHING NOTE:** sections after every 2-3 rounds - A **Final Performance Summary** when I end the session: what leverage I used well, what I left on the table, and the single biggest improvement to make before the real negotiation
Result
Scenario: User is rehearsing a salary counteroffer after receiving a job offer $8K below their target. AI plays the hiring manager.
Round 1
User: Thanks for the offer. I was hoping we could get closer to $130K given my background — I've been quoted similar ranges at two other companies I'm talking to.
Hiring Manager (AI): I appreciate that, and I hear you on the market comparisons. Our band for this role tops out at $122K without additional approval — what specifically about your background do you think justifies going above that ceiling?
COACHING NOTE: You led with "I was hoping" — that softens your ask before you've even made it. You also mentioned competing offers but didn't name numbers or companies, which makes it easy for the manager to treat it as unverified leverage rather than real pressure. Next round, be specific: cite an actual figure or credential, not a general appeal to fairness.
Round 2
User: I led the migration project that cut our infra costs by 30% at my last role, and I have a competing offer at $128K. I'd like to match that here since I'd rather join this team.
Hiring Manager (AI): That's a concrete achievement, and I appreciate you being direct about the number. I can't get to $128K without a title bump, but I can go to $124K base plus a $5K signing bonus to bridge the gap this year. Would that work as a starting point?
COACHING NOTE: Much stronger — naming the achievement and the exact competing number gave the manager something concrete to respond to, and you got a real concession (a $5K bump plus bonus structure) instead of a flat no. One miss: you said "I'd rather join this team" before they countered, which told them you're leaning toward accepting even a weaker offer. Save that signal for after you've secured the best number, not before.
Final Performance Summary: You improved significantly between rounds once you shifted from a general appeal ("I was hoping") to specific leverage (a named achievement and a real competing number). The manager's $124K + $5K signing bonus counter is a realistic outcome for this kind of gap — you likely could have pushed for the full $128K in base by holding the line one more round instead of signaling your preference for the role. Biggest improvement for the real conversation: hold your enthusiasm for the role until after the number is settled, not before.
Most negotiation prep happens one of two ways: silently, in your own head, or out loud with a friend who is too polite to actually push back. Neither exposes the weak points in your argument the way a real counterparty will. This prompt is designed to do that instead — it runs an actual multi-round negotiation roleplay, with the AI playing a realistic, self-interested counterparty rather than a cooperative scene partner.
Why this prompt is structured the way it is
The Role section deliberately asks you to specify who the AI is playing — a hiring manager, a vendor, a landlord — because a generic “play a negotiator” instruction produces generic pushback. A hiring manager protecting a fixed salary band behaves differently than a vendor protecting margin on a renewal, and the prompt needs that specificity to generate realistic resistance rather than a caricature.
The Context section's last field — your walk-away point — exists purely for your own reference, and the Constraints section explicitly tells the model never to let the counterparty see it. That's a deliberate design choice: a real counterparty doesn't know your floor either, and if the AI could see it, it would negotiate down to exactly that number every time, which defeats the point of practicing.
The coaching notes are the actual value of this prompt, not the roleplay itself. Anyone can imagine an opposing argument; what's hard to see on your own is which specific phrases undercut your position — hedging language, revealing information too early, softening a firm ask into a request. Forcing the model to break character every 2-3 rounds and name the exact moment something worked or fell flat turns a rehearsal into a diagnostic.
Where this differs from a typical negotiation prep prompt
Most negotiation prompts in circulation produce a static script — a set of talking points to memorize. This one is iterative by design: your second and third rounds should look different from your first, because you're adjusting based on real pushback and real coaching in between. That's closer to how negotiation skill actually improves than reading a list of tips would be.