Discovery Call Questions with AI (2026): Uncover Real Needs Before You Pitch

Last updated: October 2026

Priya, an account executive at a London HR-tech company, was preparing a discovery call questions document at 9pm — again. She had a 40-question list she'd inherited, a prospect who was the VP of People at a 900-person retail chain, and no idea which ten questions actually mattered. The call the next morning went fine. It always went fine. It just never produced anything she could build a demo on: lots of nods, no numbers, no decision process, no urgency.

The problem is not the questions available online — HubSpot publishes 20, RepVolt lists 50, other pages push 75. The problem is that a question bank without a framework produces pleasant conversations, not qualified pipelines. This guide fixes that: 45 discovery call questions organized by SPIN and MEDDIC stages, a 30-minute agenda that fits them, and AI prompts that generate a tailored bank from your specific ICP in minutes.

The short answer: ask 11–14 questions on a discovery call. Gong's analysis of 519,291 sales conversations found that range maximizes next-step conversion — more than 14 creates an interrogation, fewer than 11 leaves the deal shallow. Listen 46% of the time, talk 54%, and probe 3–4 distinct problems before you ever discuss your product.

The AE who stopped interrogating and started closing

Priya's turn came from a blunt coaching session. Her manager listened to three recorded calls and counted: she asked 23 questions in 30 minutes, switched topics 11 times, and never followed up on any answer. Every question was good. The sequence was noise.

They rebuilt her approach around Gong's data — trim to 12 questions, cluster them by problem, and drill down on each answer instead of moving on. She used an AI prompt (Prompt B in this guide) to compress her 40-question list into a prioritized 12 for a VP-of-People persona, and a second prompt to generate three follow-up probes for any answer she heard live.

Two months later her discovery calls averaged 13 questions, her talk ratio moved from 65/35 to 54/46, and — the number her manager cared about — deals that started with her discovery calls reached next steps 70% of the time, up from roughly half. The bank did not change. The framework did.

The question bank: 45 questions mapped to SPIN and MEDDIC

SPIN (Situation, Problem, Implication, Need-payoff) comes from Neil Rackham's research on 35,000 sales calls. MEDDIC-lite adds the qualification layer enterprise teams need: Metrics, Decision process, Pain. Use 11–14 total per call, pulled from across the stages.

Situation questions (pick 2, not 8)

1. "Walk me through how your team handles [core process] today."
2. "What's the stack behind that — and what's connected to what?"
3. "Who owns this process day to day, and who's accountable for the result?"

Keep these to two. Prospects resent re-stating things you could have Googled — and remember, HubSpot's trends data shows 96% of buyers research you before the call, so assume they notice when you haven't researched them.

Problem questions (pick 4–6)

4. "Where does this process break down most often?"
5. "What made you take this call now, rather than last quarter?"
6. "What's the workaround your team hates the most?"
7. "If you could change one thing about how this works, what would it be?"
8. "What's the complaint you hear most from [their team/customers]?"
9. "When this last went wrong, what did it cost you — in hours or dollars?"

Implication questions (pick 3–4 — where deals are won)

10. "You said reports take three days — what does your team not get done in those three days?"
11. "How does that delay affect [revenue/sales/retention] downstream?"
12. "If this stays the same for another year, what happens to your targets?"
13. "Who else feels this problem when it hits — your CFO? Your customers?"
14. "What has trying to fix this internally cost so far?"

Need-payoff questions (pick 2–3)

15. "If that number moved from X to Y, what would that make possible for you?"
16. "How would your team's week change if this ran itself?"
17. "What would make this project an obvious win internally?"

MEDDIC-lite qualification (pick 3–4)

18. "What metric will you use to judge whether this project succeeded?"
19. "Who besides you weighs in on a decision like this?" (Gartner's May 2025 research puts typical B2B buying groups at 5–16 people — always ask this.)
20. "Walk me through the last time your team bought something like this — what did that process look like?"
21. "What's the budget conversation you'd need to have, and when?"
22. "What happens if you do nothing this year?"

The disqualification questions most reps avoid

23. "Honestly — is this a priority this quarter, or a nice-to-have?"
24. "What else is competing for this budget right now?"
25. "If we're not the right fit, who would you want me to point you to?"

Asking "who else competes for this budget" early sounds risky. It is not. Deals lost to no-decision outnumber deals lost to competitors in most B2B funnels — Belkins' funnel data shows only about 12% of SQLs close-won — so disqualifying fast is a win, not a loss.

The 30-minute discovery call agenda

The summary at minute 22 is the highest-impact moment of the call. Repeat their words, not yours: "You said reporting eats three days a month and your CFO is asking why." Prospects correct you, add detail, and — because you sounded like them — trust the next step you propose. This recap also becomes the raw material for a personalized demo, which our sales demo scripts with AI guide picks up directly.

The AI question-bank workflow: 4 prompts

This is the workflow Priya used. Run it in ArWriter — the prompts are in the sales library — or any capable writing tool.

Step 1 — Generate the raw bank for your ICP with Prompt A below.

Step 2 — Prioritize to Gong's 11–14 range with Prompt B; it outputs a timed call-flow table.

Step 3 — Prepare live follow-up probes. During the call, paste the prospect's answer into Prompt C to get three drill-down questions in seconds.

Step 4 — Auto-summarize into a recap and next-step assets with Prompt D: a recap email, plus the one-page brief your demo will be built from.

PROMPT A — Generate the question bank

You are a B2B sales consultant. Create a discovery call
question bank for selling [product] to [ICP: role, company
size, industry] at [stage, e.g., Series B]. Organize by
SPIN: 4 situation, 6 problem, 5 implication (each tied to
financial or time cost), 3 need-payoff. Then add 4 MEDDIC
qualification questions covering metric, decision process,
budget posture, and competing initiatives. Phrase every
question conversationally, ≤ 20 words each.

PROMPT B — Prioritize to the proven range

Here are [40] candidate discovery questions for [ICP]:
[paste]. Select and order the best 12 (Gong's 11–14 sweet
spot) to cover: pain, cost of inaction, decision process,
budget, and competing initiatives. Output a call-flow
table with timing for a 30-minute call, and mark which 3
questions are must-ask if we run short.

PROMPT C — Live follow-up probes

Based on this prospect answer: "[paste their answer]",
suggest 3 follow-up questions in sequence: one that goes
deeper into the cause, one that quantifies the financial
or time impact, and one that identifies who else inside
the company is affected. Keep each under 15 words and
conversational.

PROMPT D — Auto recap + hand-off brief

Turn these discovery notes [paste notes] into: (1) a
recap email to the prospect summarizing their pain, the
cost of inaction, their success metric, the decision
process, and the agreed next step — in their words, not
ours; (2) a one-page internal brief formatted to feed a
personalized demo script: pain / cost / metric /
decision-maker / competing initiative.

The recap email is not a courtesy — it is a trap for misunderstandings while they are still cheap to fix. If the prospect replies "that's not quite right," you just learned something most reps only discover at contract stage.

What the research says about discovery calls

• Gong analyzed 519,291 sales conversations and found the discovery sweet spot is 11–14 questions: more than 14 and next-step rates fall, fewer than 11 and deals stay shallow. Top performers keep a talk-to-listen ratio of 54:46 — they talk slightly more than half the time — and discuss 3–4 distinct prospect problems per call, which maximizes next steps.
• Gong's scheduling data shows Monday has the highest discovery-call attendance (Wednesday is a close second), Friday is the worst, 3pm is the strongest slot, and early-morning calls produce the most no-shows.
• HubSpot's sales trends research found 96% of prospects research you before speaking — so situation questions about facts you could have looked up actively cost you credibility.
• Gartner's May 2025 buying research puts the typical B2B buying group at 5–16 people, which is why single-threaded discovery kills deals: if you haven't asked "who else weighs in," you don't have a deal, you have a pen pal.
• Salesforce's State of Sales (6th edition) found 83% of AI-using sales teams saw revenue growth versus 66% of non-AI teams, with AI saving reps about 2 hours a day — hours that used to go to writing recap emails by hand.
• Cognism's connect data (200,000+ calls) shows 65.6% of cold-call connects become conversations — discovery is the step you earn with a good cold call script, which is where our cold call scripts with AI guide starts the funnel.

Fix your talk ratio in two weeks

Gong's 46% listening benchmark is the most-quoted, least-practiced number in sales. Here is a drill to close the gap:

Day 1 — Measure. Record three calls, paste the transcripts into an AI tool, and ask: "Count my talk share vs the prospect's, the number of open vs closed questions, and the point where the prospect's answers got shorter." Most reps discover they talk 65%+ and ask 20+ questions.

Week 1 — Cut and cluster. Cut your bank to 12 questions using Prompt B. Practice delivering each question and then counting two full seconds of silence before speaking again. Silence is where prospects volunteer the real problem.

Week 2 — Drill the follow-up. For every answer you hear, force one probe (Prompt C) before moving to the next scripted question. The goal is not asking better questions — it is asking better second questions.

Reps who run this drill typically land near the 54:46 ratio within two weeks, without changing anything about their personality or product knowledge.

8 discovery call mistakes to stop making

1. Asking 25 questions. Gong's 519,291-conversation dataset says 11–14. Beyond that, you're running an interrogation, and next-step rates drop.
2. Skipping implication questions. Problems without quantified cost produce "interesting" deals that die in procurement. Always ask what the problem costs.
3. Pitching at minute five. The prospect asked one clarifying question and you delivered nine slides of features. Discovery means their words fill the notes, not yours.
4. Single-threading. With 5–16 people in a typical buying group, leaving the call without knowing who else decides is malpractice.
5. Faking research. 96% of buyers researched you — "so, what does your company do?" ends the call's credibility in four words.
6. No written recap. Memory drifts; deals die of ambiguity. The recap email is your cheapest insurance policy.
7. Never disqualifying. Pleasant conversations with no budget and no urgency are not pipeline. Ask the hard question at minute 23, not quarter-end.
8. Ending without a next step. "I'll send you some info" is a no. Propose a specific, dated next step tied to what they told you.

Build your next question bank tonight

You can hand-build this week's discovery prep in an hour, or generate it in five minutes. ArWriter runs the full chain — question bank from your ICP, prioritization to the 11–14 range, live follow-up probes, and recaps — from one workspace, with the sales prompt library ready to go. Plans start at $4.99/month. If objections are where your deals stall after discovery, pair this with our objection handling scripts with AI guide.

What to do next

1. Run Prompt A with your actual ICP — get your first tailored bank.
2. Cut it to 12 with Prompt B and block two hours of practice.
3. Record your next three calls and measure your talk ratio.
4. Recap every call with Prompt D, and feed the brief straight into your demo prep.

Discovery is one stage in a chain. The meeting starts with the cold call script that books it, and it ends when your demo converts what you learned — the demo script guide is the next link.

Frequently Asked Questions

What should I expect on a discovery call?

Expect the seller to ask most of the questions: situation, problems, their cost, your decision process, and next steps. A good call ends with a summary of your words and a specific follow-up. Expect 11–14 questions and roughly 30 minutes — and expect the seller to have researched your company beforehand.

Is a discovery call an interview?

It resembles one, but with shared stakes. In an interview only one side evaluates; on a discovery call both sides qualify. The seller assesses fit and urgency; the buyer assesses credibility and relevance. Great sellers treat it as a structured conversation, not a checklist interrogation.

How many questions should you ask on a discovery call?

11–14, per Gong's analysis of 519,291 conversations — that range maximizes next-step conversion. RepVolt's guidance lands at 8–12 well-placed questions. The key is follow-up depth: three drilled-down answers beat fifteen checklist questions asked once.

What's the difference between a discovery call and a qualification call?

Qualification checks fit: budget, authority, timeline, need (BANT-style). Discovery explores depth: what the problem costs, why now, who's affected. In practice they blend — but a call that only qualifies produces flat deals, while a call that discovers makes qualification obvious from what the buyer volunteers.

What's the difference between a discovery call and a demo?

Discovery diagnoses; a demo prescribes. On a discovery call the prospect talks roughly half the time and no product is shown. On a demo the seller shows specifically what was discovered — which is why a great demo is built directly from discovery notes. Doing discovery inside a demo wastes both.

How long should a discovery call be?

30 minutes covers a first discovery call for most B2B deals: 3 to frame, 13 to explore problems and cost, 6 to qualify, 5 to summarize, 3 to set next steps. Complex enterprise deals warrant 45–60 minutes. Gong's data shows Monday and 3pm slots get the best attendance.

What is a good talk-to-listen ratio on a discovery call?

Gong's benchmark from top performers is 54:46 — the seller talks slightly more than half the time, usually while framing, summarizing, and teaching. Most untrained reps sit at 65%+ talk time. Record three calls and count; then cut scripted questions and add follow-up probes.

How do you take notes on a discovery call?

Use a fixed template — pain, cost of inaction, success metric, decision-maker, competing initiative, next step — and capture the prospect's exact words for each. Better: record (with consent) and let an AI tool produce the transcript summary. Feeding that template into a demo-script prompt is how notes become deals.

Sources

• Gong — discovery call research (519,291 conversations) — question count, talk ratio, problem coverage, scheduling benchmarks
• HubSpot — discovery call questions guide — staged question lists and a 30-minute agenda template
• RepVolt — 50 best discovery call questions — question grouping by stage and SPIN usage
• Salesforce — State of Sales, 6th edition — AI adoption, time savings, and revenue growth in sales teams
• Cognism — State of Cold Calling statistics — connect-to-conversation rates feeding the discovery stage