Five AI Prompts That Will Change How You Look at Your Call Notes (And the Honest Catch With Each One)

Your reps know more than your reports. They walk out of shops and service centers every day carrying intel about what's selling, who's switching brands, and why a deal stalled. Most of it dies in the parking lot before anyone writes it down, and the part that does get logged sits in HubSpot where nobody has time to read it.

And right now, you can test what AI does with that intel yourself, this afternoon, for free. No budget, no pilot, no permission from IT. Just last quarter's call notes or a handful of transcripts, and a chat tool like Claude or ChatGPT.

So before anybody sells you a platform, go run these five prompts on your own data. Each one will show you something you didn't know was in there. And I'm going to tell you exactly where each one falls apart, because the limit is the most useful part. The limit is the reason a one-off prompt and an actual system are two very different things.

Grab 10 or 15 call notes or transcripts. Paste them in. Let's go.

1. The objection finder

Here are 12 of my team's call notes from this quarter. Read all of them.
Pull out every objection or hesitation a customer raised and group them
by theme. Give me a table: theme, how many calls it showed up in, and
one real quote per theme. Then flag any objection that only came up
once but sounds like it could cost a big deal. Only use what's in the
notes. Don't invent anything to fill a gap.

What you get back is the conversation your team has been having for three years, settled in about nine seconds. Price keeps coming up, sure, but so does "your delivery window doesn't work for my counter guys," and that one you didn't see coming. You've just turned a stack of notes into a ranked list of what's actually costing you deals.

The catch: it only knows the 12 you pasted. Ask it about the other 900 calls from this quarter and it can't help you, because they were never in the window. Close the chat and the analysis is gone. Next month you start from a blank box and do the whole thing by hand again. Great for a snapshot. Useless as a habit.

2. The competitor scanner

Read these notes. Every time a competitor is mentioned, pull the quote,
tell me which competitor, and whether the customer sounded positive,
negative, or neutral. Put it in a table sorted by how often each name
comes up. If you're not sure a name is actually a competitor, list it
in a separate "not sure" section instead of guessing.

This is the one that makes people sit up. You might find a competitor getting named in a third of your calls, mostly around lead times, and your reps have been absorbing that quietly without anyone connecting the dots. That's a marketing angle and a product conversation, sitting in data you already had.

The catch: the AI doesn't know your world. It doesn't know that "the Chicago guys" is what your reps call one competitor, or that a brand shows up with three different spellings in your notes, or which of two similarly named outfits actually competes with you. It catches the obvious mentions and misses the ones that take industry knowledge to spot. You have to teach it your competitors, your products, and your nicknames every time you open a new chat. It has no memory of your business.

3. The lost-deal autopsy

Here are my notes from a deal we lost. Walk through them in order and
tell me the real reason it died, not just the reason on the surface.
Flag the earliest warning sign we missed. Back every claim with a
direct quote from the notes. Where the notes don't actually say,
write "not in the notes" instead of guessing, and tell me how
confident you are at the end.

Run this on a deal that still stings and it will find the moment things turned, usually weeks before anyone admitted the deal was in trouble. The customer told you. It's right there in the week-three note. Reading it after the fact is humbling in a useful way.

The catch: it will sound just as confident when it's guessing as when it's right. If the notes are thin, it fills the gaps with a clean story that may not be true, and a clean story is exactly what you want to believe after losing a deal. It also does one deal at a time. The pattern you actually need, the thing killing deals across the whole region, lives across hundreds of these, and you can't hold hundreds in a single prompt.

4. The pre-call briefing

Here's everything we have on this account: past call notes, the last
few orders, open quotes. Give me four things: where the relationship
stands right now, anything that looks like a risk, three specific
things to raise on my call tomorrow, and one good question to open
with. If something important looks missing, tell me what to go dig up
before the call.

Hand a rep this before a meeting and they walk in sounding like they've got a steel-trap memory of an account they last touched in March. It pulls the relationship together and shows you the openings. For a team juggling a few hundred accounts, that's real time back and a better conversation.

The catch: it only knows what you paste, and pasting it is the whole job. It can't see your CRM on its own. It doesn't know about the five other people at your company who touched this account, the support ticket from last week, or the email your colleague sent on Tuesday. The intel is scattered across systems and people, and a chat window only sees the corner of it you happen to feed it.

5. The territory read

Across all these notes, what themes differ by region or product line?
Where is demand shifting and where are the same complaints clustering?
Separate the strong patterns, the ones showing up in three or more
notes, from the weak signals that are just a hunch. For the patterns
you'd want to confirm, tell me what extra data would make you more sure.

This is the prompt that feels like a crystal ball. Patterns surface that no single rep could see from inside their own territory, because no single rep reads everyone else's notes. Demand moving in one market, the same complaint clustering in another. That's the kind of thing leadership usually finds out about two quarters late.

The catch: your whole book of business does not fit in a chat window. You can do this for a slice, for one product line or one month, but the second you want the real picture across every rep and every account and the last two years, you've blown past what the tool can hold in one go. The bigger and more valuable the question, the worse a single prompt handles it.

Notice what all five have in common

Every catch is the same catch wearing a different shirt. A chat tool is impressive on a handful of documents, in the moment, for one person. It falls down the second you need it to do four things at once: work at scale, remember what it learned, know your specific world, and connect people and accounts across your whole team. Those four gaps are not a knock on the AI. That's just the difference between a prompt and a system.

Think about onboarding a new intern. You can hand them a folder and get something useful by Friday. But they won't know your competitors, remember last month's accounts, or read across the whole team's work unless you build that knowledge into how they operate. Same with these tools. The prompt is the intern's first afternoon. The system is the intern who's been with you two years and knows where everything is.

What closing those gaps actually changes

The way I always measure a win is pretty simple. Did we save the person time. Did we give them better visibility into what needs their attention. And did we build intelligence that compounds instead of evaporating when you close the tab.

Fix scale, memory, context, and connection, and the prompts you just ran stop being party tricks and start being infrastructure. The objection finder runs across every call, not twelve. The competitor scanner already knows your competitors. The territory read holds two years instead of one month. On a recent project we pulled signal from a customer's support calls at under four minutes a call, five of them in eighteen minutes, and that was a tiny sample of what they had. None of that comes from a better prompt. It comes from a system built around your data.

I'll be honest about the limits there too. When we enrich and link accounts, we're right about 80% of the time, not 100. Anybody promising perfect is selling you something.

Where to start

Don't boil the ocean. Run the five prompts this week on your own notes and see what comes back. You'll learn more about your data in twenty minutes than a vendor deck will ever teach you.

When you hit the wall, and you'll hit it around prompt three, that's the conversation worth having. We do a small signal audit, sometimes called Project Canary: send us a slice of your existing calls, notes, or HubSpot data, and in a few weeks we hand you back a real read on what's hiding in there plus a roadmap for putting it to work. No behavior change for your reps, no rip-and-replace. Crawl, walk, run. If it's useful, we keep going.

Because at the end of the day, this industry's biggest competitor is waste. All that intel your team already gathered and never got to use is waste. I'd rather help you fight that than watch another quarter of it die in the parking lot.