How To Use AI To Prepare For A Sales Call

Preparation wins sales calls, and it is also the step that gets skipped when the calendar is full. AI changes that, because the research that used to take an hour now takes ten minutes, if you know what to ask for.

The model does not run the call. It gets you into the room knowing what the buyer cares about, what they are likely to push back on, and what to ask first.

What to feed it

The output is only as good as what goes in.

Their website, meaning the homepage, the services page and the about page.

Their recent posts on LinkedIn or Instagram, which show what they are proud of and what they are pushing this quarter.

Whatever they told you, from the enquiry form, the email thread and the booking notes.

Your own notes from any past conversation with them.

Paste it in as text. When the model browses for itself, you never know exactly which pages it read or which ones it skipped. Keep anything confidential out of tools your business has not approved.

What to ask for

Ask for specific outputs, never "tell me about this company".

A one paragraph summary of what they sell, to whom, and where.

Three guesses at the problem behind the enquiry, ranked by likelihood, with the evidence for each.

The objections they are likely to raise, drawn from your own list of past objections, per how to collect and use customer objections.

Five discovery questions written for this buyer, built off the discovery call questions that matter.

Anything that looks off, such as a site that has not been updated in years or an offer that has quietly changed.

Then read it and argue with it. The model is a fast researcher, not a decision maker, and it will sometimes state a guess as if it were a fact.

Build it once, run it every time

The first time, you write the prompt by hand. The second time, you save it.

Write one standard prompt with your offer, your ideal client and your objection list built in, then paste the buyer's details underneath and run it. In Claude or ChatGPT this can be a saved project with your context already loaded, so every call prep starts from the same base. At Ignis, repeatable prompts like this run our research across every account, which is covered in how we use AI at Ignis.

Use it after the call too

Record the call, with permission. Then run the transcript through the same model.

Ask for the buyer's exact words on the problem, every objection raised, every commitment made by either side, and a draft follow up email. The follow up goes out within the hour, written in their language rather than yours.

What not to hand over

The opening. The first 30 seconds are yours, and they need to sound like a person, per the first 30 seconds of a sales call.

Judgement on fit. The model has no idea if you actually want this client.

The relationship. A buyer can tell when every message is generated, so use AI to prepare and then talk like yourself.

A ten minute routine

Two minutes to paste their details into the saved prompt. Three minutes to read the output and mark what you doubt, then three more to pick your opening and your first two questions. The last two minutes go on checking their LinkedIn yourself, because you will notice things the model cannot.

Done every time, this is the difference between a call that feels like an interview and one where the buyer feels understood before they finish their first sentence.

What this unlocks: a sales process where every call starts informed, which raises the close rate without adding a single extra lead.