# AI Prompts for Sales Teams: Research, Objections, Follow-ups, Call Prep > Source: > Published: 2026-07-26 05:49:00+00:00 # AI Prompts for Sales Teams: Research, Objections, Follow-ups, Call Prep **Short answer:** AI can’t make you better *on*a sales call — it can’t listen, build trust, or ask the follow-up you didn’t plan. What it can do is make sure you’re on the **right** call, **prepared**, and that you **learn** from the last one. That’s where the leverage is: B2B reps spend only about a third of their time actually selling, and the average win rate is roughly 21%. The prompts below attack qualification, preparation and post-call learning — not outreach volume. **TL;DR — Key Takeaways** **Score deals to kill them, not to justify them.** The deal-likelihood prompt is written to find reasons you’ll lose. A pipeline full of hopeful deals is worse than a small honest one.**B2B reps sell only ~28–34% of the time**, losing roughly 8 hours a week to admin and data entry ([2026 productivity benchmarks](#aisp-sources)).** Average B2B win rate is ~21%**, but it swings by deal size: under $50k closes at 35–45%, above $100k at 15–25%. Benchmark against your band.** Ask 11–14 targeted questions**on discovery, and talk*less*than the prospect — top reps sit near a 43:57 talk-to-listen ratio.**Debrief every call.** The post-call prompt is the compounding one — it shows you the question you didn’t ask while you can still ask it. **✔ Best for** AEs, SDRs, founders selling their own product, and small sales teams running B2B deals with discovery calls and multi-touch cycles. **✕ Skip if** You want cold email specifically (we have a separate guide), you’re in high-volume transactional or e-commerce sales, or you’re evaluating CRM and conversation-intelligence platforms rather than prompts. **On this page** ## Why doesn’t AI help most sales reps? **Because they point it at the wrong part of the job.** Most “AI for sales” advice is about generating more outreach — more emails, more sequences, more volume. But volume was never the constraint on a 21% win rate. The constraint is that reps walk into calls underprepared, keep deals alive that were never going to close, and forget what happened on the call by Thursday. The data supports the reframe. B2B reps spend only about a third of their time selling, with roughly 8 hours a week lost to admin and data entry, plus another 9–11% on prospecting research before making contact. That’s the surface area AI should attack. **What AI genuinely cannot do:** listen, read hesitation, build rapport, or ask the unplanned follow-up question that unlocks a deal. Conversation research consistently ties winning to *listening more*and asking better questions — both irreducibly human. Any guide promising AI will close deals for you is selling something. **Prompt 1** **Qualify** Score the deal honestly, kill weak ones **Prompt 2** **Prepare** Researched questions, not a script **Prompt 3** **Objections** Library built before the call **Prompt 4** **Follow up** Specific, from real notes ## The deal context block (write this first) **One reusable block, pasted at the top of every sales prompt.** It supplies what a model can’t guess — your actual differentiator, your real proof points, and crucially, why you *lose*. Loss reasons are the most under-used input in sales prompting. ``` DEAL CONTEXT BLOCK (paste at the top of every prompt): WE SELL: [PRODUCT/SERVICE IN ONE PLAIN SENTENCE] TO: [BUYER TITLE, COMPANY SIZE, INDUSTRY] PRICE POINT: [TYPICAL DEAL SIZE AND STRUCTURE] TYPICAL CYCLE: [LENGTH, NUMBER OF STAKEHOLDERS] THE PROBLEM WE REMOVE: [IN THE BUYER'S WORDS, NOT OURS] WHAT THEY DO INSTEAD TODAY: [STATUS QUO / COMPETITOR] PROOF POINTS (only use these — don't invent): [2-3 NAMED RESULTS WITH REAL NUMBERS] OUR REAL DIFFERENTIATOR: [THE ONE THING A COMPETITOR COULDN'T ALSO CLAIM. If you can't name it, say so.] WHY WE LOSE (be honest — this matters most): [TOP 3 REASONS DEALS DIE: price, no urgency, champion left, chose incumbent, procurement, built in-house...] WHO WE'RE BAD FOR: [THE CUSTOMER WE SHOULD DISQUALIFY] RULES: - Never invent a statistic, customer name, or capability. If you need one I haven't given you, ask. - Be direct. No encouragement, no sales-coach clichés. - If evidence is thin, say the evidence is thin. ``` The “why we lose” and “who we’re bad for” lines are what make the downstream prompts sharp. Without them, every model output assumes your deal is winnable — which is exactly the bias you’re trying to correct. ## 1. How do I score whether a deal will actually close? **Use a scoring prompt built to disqualify, not to confirm.** Ask a model whether a deal looks good and it will find reasons to agree — the same flattery problem that inflates every optimistic forecast. The useful version hunts for the reasons you’ll lose, and it treats *absence of evidence* as a red flag rather than a neutral. ``` [PASTE DEAL CONTEXT BLOCK] THE DEAL: Company: [NAME, SIZE, INDUSTRY] Contact: [NAME, TITLE — and whether they hold budget] Source: [INBOUND / OUTBOUND / REFERRAL] Stage & age: [STAGE, WEEKS IN PIPELINE] What's happened so far: [CALLS, EMAILS, WHO ELSE INVOLVED] What they've SAID: [QUOTES IF YOU HAVE THEM] What they've DONE: [ACTIONS: shared data, brought in a colleague, asked for pricing, booked a follow-up, nothing] Next step booked: [YES + DATE, OR NO] Score this deal. Your job is to find reasons it WON'T close. 1. EVIDENCE AUDIT — split what I've told you into what the prospect SAID (opinions, interest, praise) and what they have DONE (actions with a cost). Only the second column predicts anything. 2. THE FIVE GAPS — for each, state KNOWN, ASSUMED, or UNKNOWN: - Pain: is there a real, quantified cost of doing nothing? - Power: have I spoken to whoever signs? - Urgency: is there a reason to act NOW, not next quarter? - Process: do I know their buying process and timeline? - Competition: do I know what else they're considering, including doing nothing? 3. LOSS MATCH — compare this deal against my stated loss reasons. Which pattern does it most resemble? Quote the evidence. 4. SCORE — LIKELY / AT RISK / LIKELY DEAD, with the single biggest reason. Treat UNKNOWN as negative, not neutral. 5. THE ONE QUESTION — what should I ask next that would most change this assessment? Give me the exact wording. 6. WALK AWAY? — if this deal doesn't deserve more of my time, say so plainly and explain why. Do not soften the verdict. "Interest" is not a buying signal. ``` **Why it works:** the SAID/DID split in step 1 is the whole prompt. Most pipelines are propped up by enthusiasm — prospects who said encouraging things and did nothing. Forcing the two columns apart makes that visible in a way a stage field in your CRM never will. **▲ Likely** Economic buyer engaged. A quantified cost of inaction. A booked next step. The prospect has done something with a cost — shared data, pulled in a colleague, asked procurement a question. **◆ At risk** Champion is real but you haven’t met power, or there’s no compelling reason to act this quarter. Fixable — but only if you name the gap and go get it. **▼ Likely dead** Lots of positive language, no actions with a cost, no next step booked, no access to the decision-maker. This is the deal that sits in your pipeline for five months and then goes quiet. ### Get the Objection-Handling Prompt Library 50+ objections mapped to the real concern underneath each, with response frameworks, plus every prompt in this guide and the deal context block template. Free. [Get the library →](#aisp-lead-magnet) ## 2. How do I prep for a sales call with AI? **Generate questions, not a script.** Call outcomes correlate with asking roughly **11–14 targeted questions**, and with the rep talking *less* than the prospect — top performers sit near a **43:57 talk-to-listen ratio**. A script pushes you toward monologue. A question bank pushes you toward discovery. ``` [PASTE DEAL CONTEXT BLOCK] CALL DETAILS: Who: [NAME, TITLE, COMPANY] Call type: [DISCOVERY / DEMO / NEGOTIATION / RENEWAL] What I already know: [RESEARCH, PRIOR CALLS, TRIGGERS — job postings, funding, launches, reviews, public statements] My goal for this call: [ONE SPECIFIC OUTCOME] Prepare me. Give me: 1. THE HYPOTHESIS — based on what I know, what's my best guess at their actual problem? Mark it clearly as a hypothesis to test, not a fact to assert. 2. 14 QUESTIONS, grouped: - Situation (3): how things work today - Problem (4): what's broken, and what it costs - Impact (3): who else it affects, what happens if nothing changes - Process (2): how decisions like this get made here - Competition (2): what else they're weighing Every question must ask about SPECIFIC PAST EVENTS or CURRENT STATE. No hypotheticals. No "would you". No question that leads to my product. 3. THE ONE QUESTION most likely to change how I see this deal. Mark it. 4. LANDMINES — 3 things I might say that would damage this call, given who they are. 5. WHAT I'M ASSUMING that I should verify in the first 10 minutes. 6. DISQUALIFY SIGNALS — 3 things I might hear that should make me end this politely rather than push forward. Reminder in your output: my target is to talk less than half the time. Flag if my stated goal requires me to present more than I listen. ``` Point 6 is the one reps skip and managers wish they wouldn’t. Knowing your disqualify signals *before* the call is what stops a bad-fit prospect becoming a three-month pipeline zombie. ## 3. How do I build an objection-handling library? **Build it before the call, not during.** The goal isn’t a script to recite — it’s having already thought about the real concern beneath each objection so you’re not improvising under pressure. “Too expensive” is almost never about price. ``` [PASTE DEAL CONTEXT BLOCK] Act as a skeptical, qualified buyer: [TITLE] at a [COMPANY TYPE]. You have budget and the problem is real, but you're unconvinced and you've been pitched a lot. Our offer: [WHAT WE'RE PROPOSING, AND THE PRICE] 1. List the 10 objections you'd actually raise, ranked by how likely each is to kill the deal. 2. For each, separate: - THE STATED objection (what they'd say out loud) - THE REAL concern underneath it - THE EVIDENCE that would resolve the real concern 3. Write a response to each in this shape: - Acknowledge without agreeing or getting defensive - Ask one clarifying question to find the real concern - Offer specific evidence from my proof points - Propose a concrete next step Under 60 words each. No "I totally understand." No "great question." No reframing their concern as a benefit. 4. FLAG which objections are NOT sales problems — where the honest answer is that our product, price, or positioning is genuinely wrong for them. Say which we should fix rather than handle. 5. For the top 3, give me the exact question I could ask EARLIER in the cycle so the objection never forms. ``` **Why it works:** point 4 is where competitors’ objection lists stop. Some objections are signal, not friction — if three prospects in a row raise the same one, that’s product feedback, and treating it as a rebuttal problem means never fixing it. Point 5 turns reactive handling into prevention. ## 4. How do I write follow-ups that don’t get ignored? **Reference something only you could know.** A follow-up that restates your value proposition reads like a template, because it is one. A follow-up quoting something the prospect actually said proves you listened — which is the only thing that separates you from the four other vendors in their inbox. ``` [PASTE DEAL CONTEXT BLOCK] MY RAW CALL NOTES: [PASTE EVERYTHING — messy is fine. Include their exact words where you have them, and what you agreed.] Write the follow-up email. Structure: - Open by referencing something SPECIFIC they said. Use their words, not a paraphrase into my language. - Confirm the problem as THEY framed it, not as I'd pitch it. - One piece of evidence that speaks to their specific concern. - The next step exactly as we agreed it, with a date. Constraints: - Under 120 words. - No "great speaking with you today." - No "as discussed", "circling back", "just following up". - No recap of my product's features. - One call to action. Never two. - Only use facts present in my notes. If I didn't record something you need, ask me — do NOT invent a detail about the conversation. Then, separately: - 3 things I should have asked but didn't, based on these notes — and whether it's still possible to ask them. - The single biggest risk to this deal, from the notes. ``` **Never let a model invent a call detail.** A follow-up referencing something the prospect didn’t say is worse than a generic email — it proves you weren’t listening *and*that a machine wrote it. The “only use facts in my notes” constraint is not optional. ## What separates a good sales prompt from a bad one? | Move | ❌ Weak prompt | ✅ Strong prompt | |---|---|---| Qualification | “Does this deal look good?” | “Find reasons this won’t close. Treat unknowns as negative.” | Evidence | Accepts “they seemed really interested” | Splits what they SAID from what they DID with a cost | Call prep | “Write me a discovery call script” | “14 questions about specific past events, grouped, plus my disqualify signals” | Objections | “How do I overcome ‘it’s too expensive’?” | “What’s the real concern underneath, and which objections are product problems?” | Follow-up | “Write a follow-up email” | Raw notes pasted + “only use facts in my notes, ask if something’s missing” | Learning | Call ends, notes rot in the CRM | Debrief prompt run after every call, patterns tracked monthly | The through-line: weak prompts ask AI to *encourage* you. Strong prompts ask it to *audit* you — and treat the absence of evidence as a finding rather than a blank. ## Can you show a real prompt with its actual output? Prompt 1 (deal scoring) run on a deal that most reps would have forecast as committed. ``` CONTEXT: We sell a compliance-tracking platform to ops leads at 200-800 person logistics firms. £40-70k/yr. Cycle ~4 months, 3-5 stakeholders. Proof: cut audit prep from 3 weeks to 4 days at Kestrel Freight. Differentiator: we're the only one with pre-built templates for their specific regulator. WHY WE LOSE: (1) they decide to build it internally, (2) champion has no budget authority, (3) no deadline forcing a decision, so it slips to next year. THE DEAL: Company: Halden Logistics, 400 staff. Contact: Ops Manager, does NOT hold budget. Stage: "Proposal sent", 9 weeks in pipeline. So far: 3 calls, all with the same person. Sent proposal 3 weeks ago. SAID: "This is exactly what we've been looking for." "I'll take this to the leadership team." "Budget shouldn't be an issue." DONE: Attended 3 calls. Nothing else. Next step booked: No. Score it. Find reasons it won't close. ``` **1. Evidence audit** **SAID:** “exactly what we’ve been looking for,” “I’ll take it to leadership,” “budget shouldn’t be an issue.” All three are enthusiasm with no cost attached. The third is the weakest signal in the set — someone without budget authority speculating about budget. **DID:** Attended three calls. That’s it. Nine weeks in, with a proposal out for three of them, and the prospect has taken no action carrying any cost — no data shared, no colleague introduced, no procurement question, no follow-up requested. **2. The five gaps** **Pain — ASSUMED.** You have enthusiasm but no quantified cost of doing nothing. Nowhere in these notes does anyone say what their current audit prep costs them.**Power — UNKNOWN, trending negative.** Three calls, one contact, explicitly no budget authority. You have not met anyone who can sign.**Urgency — UNKNOWN.** No deadline, no regulator date, no event forcing action. Nothing here explains why this quarter rather than next year.**Process — UNKNOWN.**“I’ll take this to the leadership team” is not a buying process. You don’t know who’s on it, when it meets, or what it needs.**Competition — UNKNOWN.** Including the option of doing nothing, which is currently winning. **3. Loss match** This matches loss reasons 2 and 3 simultaneously — champion without budget authority, and no forcing deadline. It also carries early signs of reason 1: a 400-person logistics firm with an engaged ops manager and no urgency is exactly the profile that decides to build something internally. **4. Score: LIKELY DEAD** Biggest reason: nine weeks, three calls, one contact, zero actions with a cost, and no booked next step after a proposal has been out for three weeks. Silence following a proposal is not deliberation — it’s the deal ending quietly. Four of your five gaps are UNKNOWN, which scores negative. **5. The one question** “When leadership discussed the proposal, what did they say was the reason to do this now rather than next year?” — this tests urgency and whether the conversation happened at all. If the answer is vague, it didn’t. **6. Walk away?** Not yet, but stop forecasting it. One direct attempt to reach the economic buyer, with a specific ask for a 20-minute conversation. If that fails within two weeks, move it out of the pipeline. The cost here isn’t the deal — it’s the reporting time and attention it’s absorbing while you avoid the harder conversation. Note what happened: a deal most reps would have called “committed, waiting on their internal process” got correctly identified as dead — and the diagnosis came entirely from the gap between what the prospect *said* and what they *did*. That distinction is the single most portable idea in this guide. ## Level-up: the post-call debrief prompt This is the prompt competitors’ lists don’t have, and the one that compounds. Reps run one call and move to the next; the lesson evaporates. Run this within an hour of every meaningful call, while it’s still possible to act on what you missed. ``` [PASTE DEAL CONTEXT BLOCK] CALL: [WHO, COMPANY, CALL TYPE, DATE] MY GOAL WAS: [WHAT I WANTED FROM THIS CALL] RAW NOTES OR TRANSCRIPT: [PASTE EVERYTHING — unedited. Include the awkward parts, the vague answers, and anything that felt off.] Debrief me. Be harsh. Six sections: A. SAID vs DID — separate everything the prospect expressed (interest, praise, intent) from everything they actually did or committed to. Quote both. B. QUESTIONS I DIDN'T ASK — based on the five gaps (pain, power, urgency, process, competition), what did I fail to establish? For each, give me the exact wording to ask, and whether it's still askable now or I've lost the moment. C. THE REAL OBJECTION — what concern was underneath what they actually said? Quote the moment they hesitated, deflected, or changed the subject. That's usually where it is. D. TALK RATIO & CONTROL — from these notes, did I present more than I listened? Where did I pitch when I should have asked? Quote the specific moment I took over. E. WHAT I ASSUMED — list every belief I now hold about this deal that is NOT supported by something they said or did. These are my blind spots. F. NEXT MOVE — the single highest-leverage next action, the exact wording to use, and by when. If the honest answer is to disqualify, say so. Rules: - No encouragement. Do not tell me the call went well. - If my notes are too thin to assess something, say so — that itself is a finding about my note-taking. - Distinguish "they confirmed the problem" from "they will buy a solution." These are not the same. ``` **Why this is the unlock:** Section B is the one that changes outcomes, because it surfaces the unasked question *while you can still ask it* — usually in the follow-up email you’re about to write. Section E is the uncomfortable one: it lists everything you believe about the deal that nobody actually told you. That list is where forecast errors live. **Setup tip:** keep one persistent thread per major deal so the debriefs accumulate. Patterns across three calls with the same account are far more revealing than any single debrief — and it makes handover to a colleague trivial. ## What should AI never do in your sales process? | Never hand over | Why | |---|---| Details about a conversation | A fabricated reference to something the prospect “said” destroys trust instantly and proves the message was automated. Only use what’s in your notes. | Claims, stats & customer names | Models invent plausible proof points. In a sales context that’s a misrepresentation risk, not just an accuracy one. | Pricing & contract terms | Discounts, terms and commitments carry commercial and legal consequence. Draft internally if you like; never let generated terms reach a customer unreviewed. | The forecast | A scoring prompt surfaces gaps in qualification. It has no predictive validity. Don’t convert its output into a probability you report upward. | The relationship | Prospects can tell when they’re being processed. Draft with AI, show up as a person. | The reliable pattern: **AI prepares and audits; you listen and decide.** Every prompt here is built to make you sharper in the room, not to replace you in it. ## Which model for which task? Prompts are model-agnostic. Practical notes as of **July 2026**: | Task | Best fit | Why | |---|---|---| Account & buyer research | A model with live web search | Training data alone invents funding rounds and misses this quarter’s hires. Always click the source. | Call debrief on a transcript | Claude | Large context holds a full hour-long transcript without chunking, which matters for spotting the moment they hesitated. | Deal scoring & objections | Claude or ChatGPT | Both sustain an adversarial stance without drifting back into reassurance. | Follow-ups at volume | Whichever you can call via API | Once the structure is locked, throughput matters more than nuance — but keep the notes-only constraint. | Deal history across calls | Any model with persistent projects | One thread per account so debriefs accumulate and patterns become visible. | We re-check these notes whenever a major model ships. If you’re reading this more than two weeks after the date above, verify your model versions still match. ## Frequently asked questions ### What are the best AI prompts for sales reps? The highest-value sales prompts aren’t the ones that write outreach. They’re the prompts that prepare you before a call, score deals honestly so weak opportunities leave your pipeline, and debrief calls afterwards so you learn what you missed. Writing prompts save minutes; preparation and qualification prompts change win rates. ### Can ChatGPT help me close more deals? Indirectly. AI can’t listen for you, build trust, or ask the follow-up question you didn’t plan — and those are what win deals. What it can do is make sure you walk into the right calls prepared, that you stop spending time on deals that will never close, and that you extract lessons from calls you’d otherwise forget. ### What is a good B2B sales win rate? Published 2026 benchmarks put the average B2B win rate at around 21%, though it varies sharply by deal size. Deals under $50,000 tend to close at roughly 35–45%, deals between $50,000 and $100,000 at 25–35%, and deals above $100,000 at 15–25%. Compare yourself to your own deal-size band rather than the blended average. ### How many questions should I ask on a discovery call? Analysis of large volumes of recorded discovery calls associates success with asking roughly 11–14 targeted questions, with top performers asking more than average sellers. Quality matters more than count — questions about specific past events and current workarounds produce far more useful information than hypothetical or feature-led questions. ### How much should I talk on a sales call? Less than you probably do. Conversation-intelligence research associates top performers with a talk-to-listen ratio around 43:57 — the rep speaking slightly less than the prospect — and with frequent back-and-forth rather than long monologues. Discovery calls in particular reward a lower share of rep talk time. ### How do I use AI to handle sales objections? Use it to build a library before the call, not to improvise during one. Have the model list the objections a skeptical qualified buyer would raise, identify the real concern underneath each stated objection, and draft short non-defensive responses. Ask it to flag any objection that signals a genuine product or pricing problem you should fix rather than handle. ### Should I let AI write my sales follow-up emails? Draft with AI, send as yourself, and always personalise from your actual call notes. A follow-up repeating generic value propositions performs worse than a short message referencing something specific the prospect said. Never let a model invent a detail about the conversation — a single fabricated reference destroys the credibility the follow-up depends on. ### Can AI predict which deals will close? Not reliably, and treating its output as a forecast is a mistake. What a scoring prompt does well is expose what you *don’t* know: whether you’ve met the economic buyer, whether a compelling reason to act now exists, and whether the prospect has taken any action rather than merely expressing interest. Use it to surface gaps in qualification, not to generate a probability you then trust. ### Download: The Objection-Handling Prompt Library 50+ common B2B objections mapped to the real concern underneath each, with response frameworks and the earlier question that prevents them. Includes every prompt in this guide plus the deal context block template. [Send me the library →](#) Enter your email and we’ll send the library plus a short monthly prompt update. Unsubscribe anytime. **Written by the Narracomm team** Narracomm is a communications and content strategy team that helps business owners, operators, and founders use AI to produce clear, credible, high-performing work. We build and test these prompt systems inside real client sales processes — across qualification, discovery and follow-up — and revise them as models change. [Add specific credentials, quota-carrying or sales leadership experience, deal sizes and industries, and a named reviewer here to strengthen E-E-A-T.] ## Sources & further reading [Everstage — Sales Productivity Statistics: Trends & Data for 2026](https://www.everstage.com/sales-productivity/sales-productivity-statistics)[Salesmotion — Sales Win Rate Benchmarks 2026](https://salesmotion.io/blog/sales-win-rate-benchmarks-2026)[PipelineGrader — The State of B2B Sales in 2026: Win Rates & Pipeline Benchmarks](https://pipelinegrader.com/insights/b2b-pipeline-cac-benchmarks)[Gong — Mastering the talk-to-listen ratio in sales calls](https://www.gong.io/blog/talk-to-listen-conversion-ratio)[Gong — Discovery call techniques (analysis of 519,000+ recorded calls)](https://www.gong.io/blog/deal-closing-discovery-call)[Gong — Top objections across 300M cold calls](https://www.gong.io/blog/we-found-the-top-objections-across-300m-cold-calls-heres-how-to-handle-them-all)[SPOTIO — 140+ Sales Statistics (2026 update)](https://spotio.com/blog/sales-statistics/) Last reviewed and updated: **July 25, 2026** · Benchmarks and model notes verified against current sources. Next review due within 14 days.