Last updated: June 9, 2026
Half of the prospects who book time with your team will never buy — not because your product is wrong, but because they were never a fit to begin with. The discovery call is where you find out which half is which. Run it well and you protect weeks of selling time; run it badly and you burn a quarter chasing deals that were dead on arrival.
Yet most discovery calls are run like a form-filling exercise: a few generic questions, a rushed pitch, and an optimistic note in the CRM. The data shows how costly that is. Across 939 B2B companies, only 13% of deals that enter discovery go on to close, and the single biggest leak in the funnel comes right after discovery, at the qualification-to-proposal step. This guide breaks down how high-performing teams run discovery calls that qualify ruthlessly, surface real business pain, and set up every later stage to win — and where AI now carries the load around the conversation.
A discovery call is the first substantive conversation with a prospect who has shown real buying intent. Its job is not to demo and not to pitch — it is to diagnose. You are mapping the prospect's problems, quantifying their impact, identifying who actually controls the decision, and deciding whether the opportunity deserves more of your team's time.
That diagnosis matters because the funnel is unforgiving. In the Optifai pipeline benchmark, deals advance from discovery to qualification only about 40% of the time, and the qualification-to-proposal stage is where most opportunities quietly die. When a deal lingers in discovery longer than three weeks, it is usually a signal that the rep is chasing an unqualified lead rather than working a genuine buyer.
Discovery is also the part of selling that AI is least likely to replace. Gartner predicts that by 2030, 75% of B2B buyers will prefer sales experiences that prioritize human interaction over AI. The reps who win will be the ones who use that human conversation to the fullest — and let software handle the prep, capture, and follow-up around it.
Strong discovery calls follow a repeatable arc. The table below maps the four phases, what each one is for, and roughly how much of a 30-minute call to spend on it.
| Phase | Goal | Share of call |
|---|---|---|
| Open & set agenda | Build trust, confirm available time, align on the outcome | ~15% |
| Diagnose | Uncover 3–4 real problems and quantify their business impact | ~55% |
| Demonstrate value | Tie one relevant customer story to their exact situation | ~15% |
| Lock next steps | Calendar a specific, named next action before you hang up | ~15% |
Top reps research every prospect before the call — recent funding, org structure, tech stack, and the specific pain their role tends to feel. Walking in cold costs you credibility in the first three minutes. This is the easiest part to automate: an AI pre-call research workflow can assemble a one-page brief on the account and a tailored question set so reps spend their energy on the conversation, not the homework.
The diagnosis phase is where deals are won or lost. Two numbers from an analysis of thousands of discovery calls separate top reps from everyone else: they ask 11–14 questions — enough to qualify, not so many it becomes an interrogation — and they talk only about 46% of the time. Underperformers dominate the conversation, talking roughly 72% of the time, and learn almost nothing about the buyer in the process.
The goal of the diagnosis is to surface three to four meaningful problems, then push past the symptom to the business consequence. "Our response times are slow" is a symptom. "We lose two enterprise deals a quarter to faster competitors" is a quantified, fundable problem — the kind that justifies a budget and a decision.
The last five minutes determine whether the deal has legs. Recap the pain in the prospect's own words, confirm you can help, surface any objections while you are still on the line, and lock a specific next action with a date and the names of who should attend. "I'll follow up next week" is how deals go dark; "Thursday at 2pm, with your RevOps lead, to map the rollout" is how they advance.
Great discovery is built on a deliberate mix of question types. Situational questions set context, pain questions surface friction, impact questions attach a dollar figure, urgency questions establish a timeline, and decision questions reveal the buying committee. A few worth stealing:
Whatever methodology your team runs — BANT, MEDDIC, or MEDDPICC — your questions should ladder up to it. Teams that apply disciplined MEDDIC and MEDDPICC qualification keep stage-to-stage progression high because they only advance deals that clear every box. The point of the questions is a confident yes or no on whether this is a real, funded, winnable deal — the same judgment that powers good lead qualification earlier in the funnel.
AI does not replace the human conversation — it removes the friction around it. Before the call, it researches the account and drafts questions. During the call, real-time assist surfaces the next best question and flags buying signals. After the call, it captures notes, scores the opportunity, and drafts the follow-up.
The most direct lever is making sure reps only spend discovery time on prospects worth it. An AI sales agent such as Darwin's inbound qualification worker, Alba, can engage, qualify, and book inbound leads automatically — so by the time a human joins, the prospect already clears the bar for budget, fit, and intent. After the conversation, conversation-intelligence coaching reviews talk-time ratio, the questions a rep missed, and whether a concrete next step was set, turning every call into a coaching moment.
Used this way, AI raises the floor on discovery quality across an entire team: every rep walks in prepared, asks better questions, and follows up the same day — without adding headcount.
Discovery quality is measurable, not a gut feel. A handful of signals separate teams whose pipeline is real from teams whose forecast is fiction:
Reviewing these as a team also surfaces coaching opportunities and keeps your sales pipeline optimization grounded in what actually happens on calls rather than what reps report in the CRM.
How long should a B2B discovery call be?
It depends on deal complexity: roughly 15–30 minutes for transactional deals, 30–45 minutes for mid-market, and 45–60 minutes for enterprise deals with multiple stakeholders.
How many questions should I ask on a discovery call?
Aim for 11–14 well-chosen, open-ended questions spread naturally through the conversation — enough to qualify thoroughly without it feeling like an interrogation.
What is the difference between a discovery call and a qualification call?
They overlap heavily. Discovery is the broader diagnostic conversation about a prospect's problems and goals; qualification is the judgment you reach during it about whether the deal is worth pursuing. Good discovery always includes qualification.
How do I know if a deal is qualified to advance?
You can name the quantified business problem, the budget, the decision process, and at least one champion. If any of those is missing after discovery, the deal is not ready to move to proposal.
Can AI run discovery calls for me?
AI can run first-touch qualification and booking for inbound leads, and it can prep, transcribe, score, and follow up around human calls. The high-stakes diagnostic conversation itself still benefits from a skilled human, which is why most teams use AI to support reps rather than replace them.
Spend discovery time only on deals worth winning
Darwin's AI agents qualify, book, and prep your inbound pipeline so your reps walk into every discovery call ready to close.
See how Alba qualifies your pipeline →