What is an AI discovery call?
An AI discovery call is a first sales conversation run by a voice AI agent. The agent joins the Zoom, Google Meet or Microsoft Teams meeting a lead booked, introduces itself as an AI, asks discovery questions, qualifies the lead against your framework and gives your account executive a written deal brief.
It's different from an AI note-taker, which sits silently in a call between two humans, and from a chatbot, which qualifies by text on your website. The AI is a participant in the conversation, not an observer.
| AI discovery agent | AI note-taker | Website chatbot | |
|---|---|---|---|
| Speaks in the call | Yes | No | No (text only) |
| Runs discovery | Yes, follows your framework | No, records a human | A few scripted questions |
| Needs a rep on the call | No | Yes | No |
| Depth of qualification | Full first-call discovery | Whatever the rep asked | Shallow |
| Output | Scored deal brief + CRM fields | Transcript + summary | Chat log |
Why teams automate discovery calls
Discovery is the most expensive half hour in B2B sales. It needs your most skilled people, it can't be batched and it doesn't scale with demand. When inbound spikes, it's the first thing to break.
- 50 hrs
- of rep time to run 30-minute discovery on 100 inbound leads
- 24/7
- the hours your buyers actually submit forms, across time zones
- 1 brief
- per call, structured the same way, every time
The result is familiar. Leads wait days for a slot (see why that costs you deals). Rushed reps skip qualification questions, and AEs walk into second calls without the context from the first.
How an AI discovery call works, step by step
- 01The lead books any slot, in their time zone
From your website, a sequence or an email reply. Because the AI's calendar never fills up, there's always a slot within the hour, including nights and weekends.
- 02The AI joins the meeting like any participant
The calendar invite includes the agent as an attendee. At the scheduled time it joins the Zoom, Google Meet, Teams or Webex link. The lead doesn't need to install anything.
- 03It discloses, and asks for consent
The call opens with a clear statement that this is an AI assistant, and recording and transcription start only after the lead agrees.
- 04It runs a real conversation
Open questions, follow-ups on vague answers, natural turn-taking, and recovery from interruptions. It answers product questions from your knowledge base and logs anything it can't answer for the AE.
- 05It scores the lead live
Each answer maps to a field in your framework (BANT, MEDDIC, MEDDPICC, CHAMP, SPICED or a custom scorecard), so by the end of the call the scorecard is filled in.
- 06It hands off
A deal brief with scores and the buyer's own words syncs to Salesforce or HubSpot, the right AE is notified, and qualified leads get the next meeting booked with that rep.
The discovery call questions an AI should ask
A good AI discovery call doesn't sound like a survey. It follows a natural arc and covers the same ground on every call:
Situation
- “Give me a quick picture of how your team handles this today.”
- “What made you look into this now?”
Pain and impact
- “What's frustrating about the current setup?”
- “What happens if you keep doing it this way for another quarter?”
- “How would you measure whether a fix worked, six months from now?”
Decision and process
- “Besides you, who would need to say yes before anything gets signed?”
- “Walk me through how a purchase like this got approved last time.”
Timing and fit
- “Is there a date this needs to be live by, and what's driving it?”
- “Are you looking at other options, or considering building it yourselves?”
These map directly onto the major qualification frameworks. We compare them in BANT vs MEDDIC, and the full question sets are on our frameworks page.
What to automate, and what to keep human
| Let the AI handle | Keep a human on |
|---|---|
| First-call discovery for inbound leads | Pricing negotiation and commercial terms |
| Off-hours and other time zones | Deep technical scoping and custom solutioning |
| Consistent framework coverage | Executive and strategic-account relationships |
| Product FAQs from approved content | Unusual, sensitive or high-stakes situations |
| Scoring, summaries and CRM updates | Any lead who asks for a person |
“Won't buyers hate talking to an AI?”
This is the most common objection, and it's a reasonable one. In practice, buyers mostly object to being misled, kept waiting or asked the same questions twice. A well-run AI discovery call avoids all three:
- It's honest from the first sentence. Disclosure isn't optional. A buyer who knows they're talking to an AI judges it on how helpful it is. One who finds out halfway through just feels tricked.
- It's available now. A conversation today beats a human call next Thursday for most buyers in research mode.
- It makes the human call better. The AE starts from the brief, so the buyer never repeats themselves.
- There's always a way out. If a lead asks for a person, they get one: the next open rep slot, or a live warm transfer.
Our security and trust page covers disclosure, consent, guardrails and data handling in detail.
What to look for in an AI discovery call tool
- Natural voice under real conditions: low latency, interruptions, crosstalk and accents. Test it live, not with a recorded demo.
- Meeting coverage: joins Zoom, Google Meet, Microsoft Teams and Webex without a plugin on the buyer's side.
- Your framework, not theirs: standard frameworks plus custom scorecards with weights and disqualifiers.
- Grounding and guardrails: answers only from approved content, with no invented pricing, discounts or roadmap promises.
- Hand-off quality: structured CRM fields and quotes, not just a transcript.
- Trust by default: mandatory AI disclosure, consent-gated recording, encryption and data residency options.
- Language coverage: if you sell internationally, check that it can hold the conversation in your buyer's language.
The bottom line
AI discovery calls don't replace your sales team. They remove the capacity limit in front of it. Every lead gets a real conversation within minutes, every conversation follows the same standard, and every AE starts with full context. That's what Sortiva does: it takes the meeting and tells you which ones matter. Hear a live call, or read what an AI SDR is for the wider picture.
Frequently asked questions
Can AI do sales discovery calls?
Yes. Voice AI agents can now join video meetings, hold natural spoken conversations, ask follow-up questions and qualify leads against frameworks like BANT or MEDDIC. They work best for first-call discovery on inbound leads, with humans taking over for negotiation and complex deals.
How does an AI join a Zoom or Google Meet call?
It is added to the meeting's calendar invite as an attendee. At the scheduled time it joins the meeting link as a named participant, just like a person would, so the lead doesn't need to install anything.
Do you have to tell leads they are talking to an AI?
You should, and in a growing number of jurisdictions you're required to. Disclosing at the start of the call is also better for trust and conversion. Recording should only begin once the lead consents, following the rules in your region.
What's the difference between an AI discovery call and an AI note-taker?
An AI note-taker silently records and summarizes a call between people. An AI discovery agent takes part in the conversation itself: it asks the questions, qualifies the lead and writes the hand-off, with no rep needed on the call.
How long should a discovery call be?
Most first discovery calls take 15 to 30 minutes. That's enough to cover situation, pain, impact, decision process and timing without turning into a demo. An AI agent can adapt the length to how much the buyer wants to share.