Early in a new role, I asked the person who'd been in business development the longest what made customers choose us over the alternatives.
He talked for twenty minutes.
None of it was on the website.
He talked about calls that happened at 11 at night when something went wrong and a customer needed a decision made fast. He talked about specific port relationships built over decades, the kind you can't replicate by showing up at a trade show. He talked about a situation from three years earlier where a competitor couldn't perform and we absorbed the work without being asked, and what that meant for a relationship that had been on the fence.
The website talked about reliability, expertise, and global reach. Those things were true. They just didn't explain why anyone chose us.
This is the GTM version of the banker's box problem.
In the last two issues, I wrote about operations data living on trucks and in request queues instead of in systems AI can use. The same problem runs through go-to-market. Companies accumulate years of signal about what customers care about, how deals actually close, and what makes them win when they should lose. That knowledge lives in the heads of the people doing the selling.
When those people leave, the knowledge leaves. When you hire new reps, they start from scratch. When you try to run AI on your sales and marketing systems, the agents produce output that sounds like your website: technically accurate, completely generic.
Most companies implementing AI on GTM right now are pointing those systems at the CRM and the website. The CRM is a system of record. The website is the official story. Neither one contains what actually closes deals.
The knowledge that closes deals falls into a few categories, none of which are captured by default.
There's customer intelligence: what customers were worried about before they bought, the language they used before they found your language for it, what almost made them choose someone else. That gap between the problem they think they have and the problem they actually have is usually where the real pitch lives.
There's competitive reality: which competitors show up in which deals, why you lose to each one honestly, the specific moments when you win and what actually made the difference. Not the matrix in the deck. The version your best reps actually carry.
And there's institutional context, the earned credibility competitors can't replicate. A company's age, the markets it survived, what it's done that most companies in the category haven't. At a company that's been operating for decades, that context is enormous. It almost never ends up in the pitch.
Getting this out of people's heads takes about four weeks. Rep interviews: not asking what they think, asking what they actually say. Win/loss reviews that go past the sanitized version. Customer conversations designed to surface how customers describe you to someone who hasn't heard of you.
That last question is worth the price of admission. Customers explain you better than you explain yourself.
What comes out is specific: the phrase that unlocks a conversation, the proof point that makes a CFO stop and listen, the competitive argument that actually lands in a late-stage evaluation. The job is to get it out, structure it so an agent can use it, and stop running AI on the official story.
The same logic holds here as it does for operations data: the knowledge exists. The question is whether it lives in a system or in someone's head.
For most companies, it's in someone's head. That person is probably your best rep, or the business development lead who's been there for thirty years. The institutional knowledge is intact. It's just not portable.
Getting it out is the invisible work that happens before the AI layer. It produces no demo, impresses no board, and takes longer than anyone wants it to. It's also the work that makes the AI sound like your company instead of every other company in the category.
The Foundation Letter publishes monthly. One idea from inside real transformations, not from a conference stage.
P.S. The Foundation-First Playbook covers the full audit process across all eight domains that determine AI readiness, including the GTM layer. Founding member price is $147 through the first 30 copies (full price $247). foundationfirsthq.com/the-playbook.html