← All Posts

The Intel That Never Leaves the Sales Call

A salesperson spends an hour with a prospect and comes away knowing an enormous amount about them.

Not the deal terms. Those go in the CRM. Everything else.

What the person is trying to build. What they're measured on. What went badly the last time they tried something like this. Whether they're the one who'll be blamed if it doesn't work. Occasionally something that never appears in a business conversation at all, like the fact that their position is up for review and they need a visible win before then.

All of that gets used exactly once, in that hour, to close the deal. Then it's gone.

What a Sales Call Actually Is

If you've never sold, you might picture a sales call as a product demonstration with an order at the end. That's not what a good one is.

The job is to get another human to talk about what they're trying to accomplish and what they're worried about, and then connect those things to what you sell. You can't do the second part without the first, which is why good salespeople are relentless about it.

So the call generates an enormous amount of context as a byproduct of the work. The salesperson uses it in the moment, to reflect the person's goals back to them and make the product feel like the route to those goals.

Then the deal closes and the context evaporates. Not because anyone decided that. Because nothing exists to catch it.

This Isn't a Salesperson Problem

Worth being clear, because the easy version of this argument blames reps and the easy version is wrong.

A salesperson in a live conversation should not be maintaining a mental list of things to type into a system afterward. That's a terrible use of their attention, it competes directly with the listening that makes them good, and it depends on memory at the end of a day full of calls.

What follows is predictable. Something material comes up on a call, doesn't reach the CRM, and never reaches whoever takes over next. Then weeks later the customer asks why an obvious thing wasn't accounted for, and the honest answer is that they already told somebody. They did. There was just nowhere for it to go.

The customer isn't wrong to be annoyed. From where they sit, they said it out loud to your company and your company forgot.

Everyone Records the Calls Already

Here's what makes this genuinely strange in 2026.

Almost every company already records and transcribes sales calls. Several tools are running on the same meeting. Summaries get generated. Action items get distributed.

And almost nobody goes back into that material for anything other than a post-mortem after a customer blows up, usually to check whether a rep promised something they shouldn't have.

The transcripts are sitting there. The context is in them. Nobody reads them.

Why AI Hasn't Fixed It

You'd think this is exactly what a model is for, and it is. But point one at a sales transcript with no further instruction and you'll get a competent summary, next steps, and objections raised. All accurate, and none of it is what we're talking about.

It gives you that because that's what you asked for. Which brings up the thing underneath this whole problem.

You have to know what's in there before you can go looking for it.

Nobody prompts for what the customer said about looking good or looking bad, because it hasn't occurred to them that a sales call contains that. Nobody asks for the things mentioned in passing that will wreck the first ninety days, like a key person going out on leave, because nobody has ever framed a transcript as a source of that.

The capability has been available for a while. The question hasn't.

The Wrong Thing to Measure

There's a related failure I've watched up close, at a company that considered itself as AI-forward as they come.

They ran every sales call through a model and scored it. Out of ten. Consistently eights and nines, and the scores were treated as a genuine signal that things were working.

Meanwhile customers were failing. Not churning quietly, failing loudly, because they weren't getting results.

The calls were good. The scoring was accurate. It was measuring how well the conversation was conducted, which correlates with almost nothing you care about six months later. All that machinery pointed at a transcript, and it produced a grade instead of the intel.

What to Actually Pull

Three things, and they take one prompt each.

The human context. What the person said they're trying to accomplish, what they're measured on, what makes them look good and what makes them look bad. This is what lets whoever picks up the account next speak to them like somebody who was in the room.

The red flags mentioned in passing. Constraints, absences, dependencies, the thing they said offhand about a system they haven't decommissioned. These are the items that surface at week six as surprises and were never surprises.

The language that caps the relationship. Anything that installs an expiration date, and anything that got given away to close, since including something is a discount and both shape what the account can ever be worth.

The Point

A salesperson already did the hardest part of understanding your customer, and did it at the moment the customer was most willing to talk.

That work is recorded. It's transcribed. It's sitting in a folder nobody opens.

You don't need a new process to capture it. You need somebody to know it's there.


Lincoln Murphy formally named and popularized Customer Success starting in 2010 and has spent 15 years connecting it to expansion revenue and commercial outcomes. Read The Premise.

Access the 5x LTV Case Study.

See how one CRM SaaS drove 5x LTV in 90 days. Full framework, milestone breakdown, and cohort analysis.

← Previous
A Discount Isn't Only a Price Reduction
Next →
“But We Have a Lot of Churn”