Chapter 7
The First Believer
The Call That Didn't Feel Like One
Karen Walsh comes onto the Zoom already talking.
"I've got seventy-five minutes, which is more than the twenty you asked for, so use it. And I want to say why I wrote to you, because it isn't flattering to my industry."
"Please."
"The post about the loan officer who can't remember faces. There was no call to action on it. You weren't selling me anything at the end, which is now so unusual that it read as a decision." She shrugs. "It also read like somebody who'd actually talked to a loan officer that week."
"I had. That morning."
"Right. So." She pulls a legal pad toward her. "You said you'd learned something that was about my problem."
Marcus has a slide deck open in another window. He does not share his screen.
"Can I ask you some things first? I'd rather not tell you about your own business."
"Go ahead."
He asks how many officers. Two hundred and six. He asks what happens when a past borrower comes back — do the officers know? Sometimes. Depends who's watching the pipeline that week. He asks how they measure retention, and gets a number that turns out on inspection to be a different number wearing retention's name.
Then he asks what they've tried.
The floodgates open, which is what happens when nobody has asked in a while.
Birthday cards that go to an address the borrower moved out of. An annual check-in call the officers dread and the borrowers find strange. And a CRM, bought two years ago, that was supposed to automate all of it.
"Nobody uses it," Karen says.
"Why not?"
"Because it sends generic garbage. And before you ask — yes, we turned on the AI writing thing last year. It's better garbage now. Grammatically it's excellent." She turns her pen over. "My officers read the draft it makes and say that doesn't sound like me, I'm not sending that. And they're right. It doesn't. So it sits there, and we pay for it."
Marcus writes that down and underlines it and doesn't say anything for a second.
"Can I tell you the thing I found?"
"That's why I'm here."
So he tells her what Boise told him — that the borrowers who leave don't scatter, that they land in a small number of the same places within weeks of a rate move, and that the reason is not that they went shopping.
"They buy the trigger the day the credit gets pulled," he says. "Your customer is getting called by somebody who knows the loan better than the officer who wrote it, and knows it first."
Karen is quiet for long enough that he checks whether the call has frozen.
"We've been treating this as a marketing problem." She isn't talking to him now. "We keep buying tools to send nicer messages. Nobody in the building has framed it as somebody actively taking them."
"I don't think most of the industry has. I could be wrong. I've got five lenders' worth of data and a lot of it is small."
"You're not wrong." She rubs her eyes. "Okay. What do I do about it?"
He does not share the screen.
"I don't know yet. Not for you specifically. I know what I'd want to look at first — whether you can see it happening at all. Most lenders can't. If somebody is picking off your book weeks after a rate move, the first place it shows up is a number nobody on your team owns."
"And how would I find that out."
"Tom — a guy who's been advising me — made me write a five-question assessment back in February. It was fine. I've been rebuilding it since, because five questions turned out to be about three too few." He hesitates. "It's twenty now, with a scoring sheet. There's no product in it. There's not even a logo."
"Is it a demo in a trench coat?"
"There's nothing in it that requires me."
"Then send it."
They talk for another half hour about things that will never appear on an invoice — her officers, a merger she survived in 2019, why the good ones leave. When she finally looks at the clock, the seventy-five minutes are gone.
"I have to go," Karen says, and then stops with her hand near the button.
"I've been pitched by a dozen vendors this year. Most of them by a bot." She counts it off, apparently to herself. "One of them congratulated us on an acquisition we didn't make. You're the first one who actually understood our business before trying to sell me anything."
Marcus doesn't have a response ready, which is probably the correct response.
"This is the best vendor call I've had in years," she says. "Except it didn't feel like one."
Consultation, Not Convincing
He asked for twenty minutes. She gave him seventy-five and then apologized for the hard stop.
He didn't pitch, didn't demo, didn't manufacture a next step. He asked questions, listened, and told her the truest thing he had, including the part where his sample is small — and the small part is what she believed.
When you actually know the field, a sales call stops being persuasion and becomes diagnosis. You aren't overcoming objections; you're helping somebody see their own situation more clearly than they could an hour ago. The person who does that becomes, by the end, the obvious person to call.
Nothing is closed. Karen has to decide whether the software solves the problem, take it to a team, and survive a budget cycle that has nothing to do with him.
But she isn't evaluating a vendor who interrupted her. She's evaluating somebody she went and found.
The Web and the Bait
One conversation is not a pipeline. The work is turning conversations into opportunities without becoming the thing Karen described in her first ninety seconds.
Build a web, not a funnel. A funnel assumes people move through stages in order, on your schedule. Buyers don't, in my experience and in every seller's I've compared notes with — they appear, engage, go quiet for a season, and resurface the week their own board asks them a question. I have no study for that and I'd treat anyone who offered you a number for it with suspicion, because the honest version is that nobody can see the quiet part.
So build threads instead of a chute, and count them. Karen's path had five before she ever wrote to him: a post she read, a second post she read, a newsletter she didn't subscribe to but saw quoted, a comment thread she lurked in, and one piece of writing that named her problem out loud. Any one of those alone does nothing. Together they made a stranger feel like a known quantity, which is the only thing that makes a cold message from a buyer possible.
Threads hold when any one of them would snap.
The machine runs the web. Expertise sets the bait. Be exact about the division, because getting it backwards is how a one-person operation loses. Scheduling, reminding, resurfacing the thing you wrote eight months ago that answers what somebody asked this morning, noticing that a reader who went quiet in March is back — that is tedious, genuinely valuable, and precisely what software is for. Doing it by hand is how solo operators lose deals to their own admin.
But the threads themselves are made of something, and the something is the part nobody can rent.
The bait qualifies, or it isn't bait. A giveaway works when it solves one specific problem immediately, proves you know something without emptying you out, and sorts people by who bothers to want it. Marcus's assessment works for a plain reason: it isn't ten tips for lenders, it's a tool that hands somebody a number about their own company. Which makes it worthless to anyone who doesn't have that company's problem — and the worthlessness is the qualification.
CRAFT, and which letters still cost something.
Consistency is nearly free now, so it has stopped being evidence. Automate it. Don't congratulate yourself for it.
Reciprocity — leave people better off whether or not they buy. This one got harder for an odd reason: it is cheap to fake at scale, the fakes flooded it, and everyone has learned to discount it. Which means doing it for real, with a small number of people, now reads louder than it did when it was rare.
Authority is expertise, and it is the only letter that can't be manufactured, rented, or generated. It is the ground the other four stand on. Take it away and the rest is a technique — and Karen can hear a technique in ninety seconds.
Friendship. Not a tactic and not a euphemism for rapport-building. It means treating somebody as a person who has a quarter, a boss, and a bad week. A model can retrieve every fact about Karen that exists in writing. It cannot notice that she sounded tired, and it would not do anything differently if it could.
Testimonials are the letter that got harder, not easier, and anyone telling you otherwise is selling review software. Third-party proof is now among the cheapest things to fabricate, which means the proof itself has stopped carrying weight — what carries weight is a named person with something to lose by being wrong about you. That is the same test as everything else in this book. It just costs more to satisfy than it used to.
Knowing when it's real. You don't ask for the sale. You notice that somebody has already decided to have the conversation: they ask about implementation, raise budget or timing unprompted, bring a colleague in, or want to know who else you've done this for.
Then make the next step small, and let them say yes. Manufacturing that moment early is how a person who was going to buy becomes a person being sold to.
The Pipeline
He sends the assessment that afternoon — day four of the new ninety, which he is counting, because Rachel will be.
Twenty questions, a scoring sheet, actions by band. No logo. Nothing in it that requires him.
Karen replies Wednesday.
"Ran it myself before showing anyone. We scored 6 out of 20, and three of my answers were 'I don't know,' which I think is the actual finding. Ops lead has it Monday. Can we do a working session after — not a demo. I want to go through what we find."
They put it in the calendar for the following week.
It isn't a sale. What it is, is a customer doing homework she assigned herself.
The inbox has stopped being quiet. Mike Donovan writes again from First National — careful, still committing to nothing, but the mail is long and specific and ends with a question about whether the assessment holds up at their size. The two mid-size conversations from before the deadline both come back to life in the same week, which Marcus suspects is the newsletter doing work he can't see. Three smaller lenders arrive over ten days: one from a reshare of the survival post, two off the list, having read for weeks without saying a word.
Then the one he reads twice.
The Texas Mortgage Bankers Association wants him for a breakout at the spring conference. Forty-five minutes on client retention. Somebody on the programming committee has been reading his posts and describes him, in the email, as a fresh voice on a stale topic.
He writes back yes inside four minutes.
It is only afterward, closing the laptop, that two facts arrange themselves in the correct order. He has never spoken at a conference. And he has attended four of them this year in a browser tab, alone, past six in the evening, which is not the same as having been at one.
He sends Rachel three sentences: how Karen found him, how Mike arrived through a comment, how the conference invitation came from a post rather than an application.
Her answer lands in under ten minutes.
"Good. None of it counts until one of them signs. Which one is closest?"
He doesn't flinch at it, which is new. He knows exactly which one.
"Karen. Working session next week."
Then he sits for a while with the other thing, the one he said yes to in four minutes, and starts a document titled TMBA — 45 min and looks at it for a long time without typing anything.
Endnotes
No citations here, and less need for them than usual: the chapter is mostly two people comparing what they have actually seen, which is the only evidence either of them has.
Three things a reader should hold loosely. The claim about how buyers really move — appearing, going quiet, resurfacing on their own schedule — is my judgment from watching it happen, not a finding. The first draft of this chapter supported it with a multi-touch statistic; I cut the number because I could not source it, and I have deliberately not replaced it with a vaguer one. Second, Karen's account of an AI writing assistant bolted onto a CRM that loan officers still refuse to use is a composite, drawn from a complaint I have heard often enough from lenders to consider typical — again, my judgment, not a measurement.
Third, and most important: the trigger-lead practice is real. Credit-bureau prescreen "trigger leads," sold when a lender pulls a borrower's credit, are an established and much-argued-about part of mortgage lending, and I state it here as fact rather than judgment. Chapter 6's note asserts the same thing more briefly; neither note is a citation, and a reader who wants one should look up the practice by name rather than take my word for it.
Karen, her company, her officers, her score, and the conference invitation are invented. The association is not.
Draft completed: December 2025 | AI-era rewrite: August 2026