How to prepare a sales discovery call with AI

Sort what you know from what you assume

Most weak discovery calls start with a confident guess. You read a hiring announcement, decide what it means, and spend the call steering the prospect toward that conclusion. Before you ask AI for anything, write two short lists: what you have confirmed from a public source or a previous conversation, and what you only suspect.

This step matters more with AI than without it. A model given a vague description will produce a plausible account of the company's challenges, and plausible is the problem. If you paste only what you can stand behind and label the rest as unknown, the plan will turn your gaps into questions instead of filling them with invention.

Known: expanded into two new countries; three open support roles on the careers page.
Suspected: the support team is under pressure.
Unknown: current tools, budget owner, whether call quality is a concern.

Ask for questions that reveal something

Ask for an agenda rather than a list of questions, and ask what each question is meant to reveal. A question with a stated purpose is easier to judge: if you would do nothing differently whatever the answer, you can drop it. The discovery call template asks for this reasoning and for a follow-up to each question, because the useful part of a discovery conversation is often the second answer.

Then cut. A generated plan tends to be complete, and a 30-minute call cannot be. Choose the three or four topics you must leave with, usually the problem, its impact, how the decision gets made, and who else is involved. Keep the remaining questions as a reserve for when the conversation goes faster than you expected.

Rehearse the objections you expect

If you have spoken to similar buyers, you already know what you are likely to hear: not this quarter, we built something internally, we need to involve IT. Take the one you find hardest and prepare for it before the call. Paste the wording as you have actually heard it, along with what you know about the deal.

Do not memorize the replies. Their value is in showing you several possible meanings of the same sentence and a clarifying question for each. ‘Not a priority’ can mean no budget, no sponsor, or no belief that the problem is real, and each one leads to a different conversation. In the meeting, ask before you answer.

Keep the call yours

Leave the plan in the background during the conversation. If the prospect raises something more important than your next question, follow it. Take notes in their words where you can, and write down what each side agreed to do next. Those notes are the material everything after the call depends on.

If you want to use AI to summarize a recording or transcript, check your company's policy and tell the participants, and get their consent where it is required. Remove names and sensitive details from anything you paste into a tool that is not approved for that data.

Follow up with something the prospect can use

After the call, give AI your notes and ask it to separate what was said from what you are inferring. Compare the result with your memory of the conversation. If the summary claims the prospect agreed to something you cannot point to in your notes, remove it before it reaches an email.

Build the follow-up around one thing you promised or one question they asked: an example, an answer, or an introduction. A message that recalls the conversation precisely and delivers something useful is worth more than a polished recap. The follow-up template produces several variations and a rule for when to stop writing, so a silent prospect does not turn into an endless thread. Read the draft before sending it, because you are the one accountable for every claim in it.