Discovery call questions that qualify (and book the next step)

A discovery call is an interview with one design rule: every question earns its place by filling a field that someone downstream will actually read. Design the set backwards from those fields and the call stops feeling like a checklist, because each question has a reason you can say out loud. Design it forwards from a template and you get the interrogation everybody complains about — twenty questions, no structure, and a set of answers nobody records. The mechanics below are the backwards version, plus the arithmetic, the recording law and the exit script.

The test every question has to pass

Before the question list, the filter that produces it: will I write this answer down, and will someone read it?

That single test removes about a third of the questions on most discovery templates. "What keeps you up at night" fails it — there is no field for the answer and no colleague who will act on it. "How does a purchase like this get approved here" passes, because the answer changes what you do next week.

The test also fixes the pacing problem. Reps ask too many questions because they are afraid of missing something. Given a filter, the set shrinks to something a person can hold in their head, which means they can listen instead of reading. A rep reading questions off a screen is audible, and it is the single biggest reason discovery calls feel like interrogations.

Two corollaries worth stating plainly. A question you ask and do not record is a cost you imposed on the prospect for nothing. And a question whose answer you could have found before the call — headcount, funding, the tech they publicly use — is worse than nothing, because it tells them you did not prepare.

Design the set backwards from the fields the next stage reads

Start at the end. What does the next conversation need to exist?

For most B2B sales the next stage — a technical call, a demo shaped to something specific, a proposal — needs five things. So you need five field groups, and every question maps to one of them.

• Trigger. What changed, and when. This is the field that predicts whether the deal moves, and it is the one most often left empty.
• Problem and current approach. What they do today, in enough detail that you can describe it back. Not "they have no system" — what the workaround actually is.
• Cost. The problem expressed in a unit the buyer already tracks: hours, headcount, missed revenue, error rate, churn. In their unit, not yours.
• Decision shape. Who is involved, what the sequence is, and what has to be true for it to proceed. Not a single name.
• Consequence of no change. What happens if they do nothing for six months. If the answer is "nothing much", that is a real answer and it should be in the record.

Five groups, and a defensible question set is roughly four questions per group. That is twenty questions, which is far more than you will get through, and that is the point — you are choosing from a set, not working through a list.

In Autocloz these live as custom fields on the lead, and the write semantics matter for incremental discovery: a PATCH to a lead's custom fields shallow-merges, so writing one key updates that key and leaves the rest intact, and sending a key as null removes it. That is what makes it safe to fill the cost field on call two without erasing the trigger note from call one.

The four question forms, and what each is good at

Question form changes what you get back more than question topic does.

Open questions produce material you did not anticipate. "Walk me through how that works today" is the workhorse of discovery, and its strength is exactly its weakness: you cannot compare the answers across twenty calls.

Closed questions produce comparable data. "Is that a monthly process or a weekly one?" You need these for the fields, and a call made entirely of them is an interrogation. Use them to pin down an open answer, never to open a topic.

Hypothetical questions test priority without asking about budget. "If this were fixed tomorrow, what would your team do with the time?" A vague answer here usually means the problem is real but small.

Negative questions are the most under-used and the most diagnostic. "What have you tried that did not work?" and "What would make you decide not to do anything about this?" People answer negative questions with unusual honesty, because they are not being sold to, and the answers frequently contain the objection you would otherwise meet in week six.

The sequencing rule: open to find the territory, closed to fix the fields, hypothetical to test priority, negative to find the blocker. Repeat per topic rather than running all the open questions first.

The opening ninety seconds decide whether you get answers

A discovery call has an agenda problem. They agreed to a conversation and they do not know what shape it takes, so they arrive defensive and give short answers until they know what is happening.

Fix it explicitly in three sentences. Say how long it will take and mean it. Say what you want to understand — "I mostly want to understand how you handle X today, and whether what we do is even relevant". Say what will happen at the end — "and at the end I'll either tell you it is not a fit, or suggest a specific next step". Then ask permission to ask questions.

That last promise is the one that changes the call. Telling someone up front that "not a fit" is a possible outcome removes the pressure that makes people guard information. It costs nothing, because you were going to disqualify them anyway if it was true, and it materially changes the quality of what you get told.

Then the first real question, and it should be the trigger question rather than the problem question: "What made this worth a conversation now, rather than three months ago?" The answer separates a dated driver from general curiosity inside fifteen seconds. If nothing in particular happened, you are on an education call, and it is worth knowing at minute two rather than minute twenty-eight.

Twenty-two questions, grouped by the field they fill

Choose from these; do not work through them.

Trigger.

• What made this worth a conversation now, rather than three months ago?
• Who raised it internally, and what prompted them?
• Has anything changed in the team or the targets in the last quarter?
• Is there a date anything is tied to — a renewal, a launch, an audit?

Problem and current approach.

• Walk me through how this works today, from the first step.
• Who actually does that work, and how much of their week is it?
• What part of it breaks most often?
• What have you already tried, and what happened?

Cost.

• Roughly how many hours a week does the workaround take?
• What does that cost you in something you already measure?
• What happens downstream when it goes wrong — who else is affected?
• If it were fixed, what would that team do instead?

Decision shape.

• When something like this goes ahead here, what does that usually look like?
• Who else has an opinion on it — not who approves, who has an opinion?
• Has anyone here bought something like this before, and how did that go?
• Whose budget would it come out of, and does that budget have a cycle?

Consequence of no change.

• If you do nothing for the next six months, what happens?
• Is there anyone pushing for this to happen, or is it in the "would be nice" pile?
• What would make you decide not to do anything about this?
• What else is competing for the same attention this quarter?

Two closers.

• What have I not asked about that I should have?
• Based on what you have heard, what would you want to see next?

That last pair is worth more than most of the middle. "What have I not asked" routinely surfaces the constraint the rep would never have guessed, and "what would you want to see next" hands the prospect the next-step design, which makes them likelier to attend it.

Quantifying the problem: the arithmetic that turns "it's a pain" into a number

Cost is the field most often left as an adjective, and an adjective does not survive an internal conversation you are not in. Do the arithmetic on the call, out loud, and let them correct it.

Illustrative worked example, with values chosen to show the method rather than to report anyone's results. Suppose a six-person sales team says each rep spends roughly forty minutes a day assembling lists and copying details between tools. That is four person-hours a day. At an assumed loaded cost of 1,200 rupees per hour, that is 4,800 rupees a day, and across 22 working days, about 105,600 rupees a month.

Now the part that makes it real: say the number back and invite correction. "So if I have that right, it is on the order of a lakh a month in time, before we talk about the meetings that do not get booked because the time went there. Does that feel about right, or am I overstating it?"

Three things happen. They correct the number, which makes it theirs rather than yours. They frequently add a second cost you did not know about. And you now have a figure in a unit their finance team recognises, which is the form the number has to be in when they repeat it to somebody else without you in the room. If you want to run the same arithmetic on the acquisition side, the CAC and LTV calculator does the version that finance will actually check.

The discipline is to keep it in their unit. Converting their hours into your pricing on the call is a pitch, and it is a pitch delivered at the exact moment they were being candid with you.

The disqualification script, and why it is the most valuable one you own

Most discovery calls should end in a disqualification, and most reps avoid saying so because it feels like losing.

The script is three sentences. "Based on what you have described, I do not think we are the right fit." Then the reason, specifically: "you need something that does X, and we do Y." Then something useful: "the people who do X well are usually A or B — want me to point you at the right one?"

Ninety seconds, and it ends better than a vague "let me follow up next quarter" that costs both of you three more calendar entries. It also produces referrals at a rate that surprises people, because you are the one seller who told them the truth and did not try to force a fit.

The test for whether you should disqualify: if you cannot name the specific thing about their situation that makes your product relevant, you do not have a deal, you have a hope. Write the disqualification reason into the record — an unrecorded disqualification means the same lead re-enters a sequence in four months and the same call gets made again by someone else.

Autocloz's free plan covers 5 users and 10 mailboxes with the booking pages, calendar sync and video links included and no per-seat fee on any of them — start free if you want the discovery notes and the next meeting on the same record.

Recording the call: one-party, all-party, and what changes in India

If you record, know the rule before the call rather than after it.

Federally in the United States, 18 U.S.C. 2511(2)(d) provides that it is not unlawful for a person not acting under colour of law to intercept a communication "where such person is a party to the communication or where one of the parties to the communication has given prior consent to such interception", unless it is done for a criminal or tortious purpose. That is the one-party consent rule.

Several states go further. California Penal Code 632 makes it an offence to record a confidential communication without the consent of all parties, and a first violation carries a fine of up to 2,500 dollars per violation. "Confidential" turns on whether a party had an objectively reasonable expectation that nobody was listening in, which on a scheduled business call is arguable in both directions — and arguable is not a position you want to be in.

In India, the Digital Personal Data Protection Act, 2023 governs personal data with a consent-and-notice architecture of its own, and a recorded call containing a named individual's voice and business details is personal data.

The operationally safe rule that satisfies all of the above without needing a jurisdiction lookup per call: announce the recording in the first fifteen seconds, in plain words, and wait for an audible yes. "I record these so I do not have to take notes — is that all right with you?" It takes four seconds, it is honest, and the answer is on the recording.

Booking the next step before you hang up

A discovery call that ends without a specific next step has produced information and no momentum, and momentum decays faster than information.

Three rules. Propose something concrete rather than a category — "a thirty-minute session with your ops lead where we walk through the import path", not "a demo". Put it in the calendar during the call rather than promising to send times; the calendar handshake and the reminder offsets that follow are covered in the B2B appointment setting guide. And name who needs to be there, using the list you collected in the decision-shape questions, because a second call with the same single person is usually a stall wearing a meeting's clothes.

If the honest next step is "nothing yet", say that and set a date to revisit that is tied to their trigger rather than to your quarter. Someone whose contract renews on 31 March wants a conversation in January, and diarising that on the call is worth more than a nurture sequence — the long-form version of that is in lead nurturing strategies.

Write the notes before the next call starts, not once the day is over. The half of an answer you still remember at 6pm is the half that confirmed what you already believed.

What a discovery call cannot establish, and what Autocloz does not do

A discovery call cannot verify anything you were told. Every answer is a claim by a person who may be optimistic, may be guessing about a colleague's budget, or may be managing you. Recording whether an answer was stated by them or inferred by you is the cheap fix, and the qualification framework people usually reach for here is examined in what is BANT.

It cannot find the person who will kill the deal. Security, procurement and legal typically appear late and were mentioned by nobody. Ask "who could stop this?" separately from "who approves it?", because they are different people and only one of them gets volunteered.

It cannot fix a call you should not have booked. If the person has no trigger and no problem, the best discovery technique available produces a well-documented no. Qualification before the call is upstream, and it is cheaper.

Autocloz does not record or transcribe calls. The notetaker ingests a transcript you paste, up to 100,000 characters, and returns a summary, action items and a drafted follow-up. It does not join meetings, capture audio or produce a transcript. If you want one, that comes from your meeting tool.

The notetaker's version 1 extraction is deterministic, not a language model. It splits the transcript into sentences, takes the first one or two as the summary, and selects action items by matching cue phrases — "we'll", "agreed to", "next step", "by Friday" and similar. That makes it predictable and auditable, and it means an implied commitment phrased without a cue phrase is not picked up. An upgrade to the AI providers is documented as a follow-on rather than shipped.

Autocloz ships no discovery-call template and no qualification object. You get custom fields with shallow-merge semantics, notes on a lead timeline, tags and a pipeline with stages you define. The question set and the field convention are yours to impose, and conventions drift unless someone owns them.

Lead scoring cannot read your discovery fields. The scoring engine has seven rule types — title, industry, company size bucket, country, tag, recent engagement and an email-verification threshold — and none of them is a custom-field predicate. A discovery answer influences a human's judgement, not a number.

If you are comparing this against a conversation-intelligence platform that records and analyses calls as its primary function, the Autocloz and Gong comparison is the honest account of what each one is actually for.