What They Cannot Copy

THE SUNDAY STRATEGY

A guy I've known for years builds custom quoting software for contractors. Good business, eleven people, profitable since 2019.

In March a customer told him, apologetically, that one of their people had rebuilt about sixty percent of what his product does in a weekend using an AI coding tool. Not as good. Not supported. Missing the edge cases that took him four years to find. But close enough that the customer was asking the question out loud.

He called me sounding like somebody had died.

Here's what I told him, and it's the thing worth thinking about on a Sunday. The sixty percent that got rebuilt in a weekend was never his business. He just thought it was, because it was the part he was proudest of.

There's a sort worth doing on everything you sell, and it goes like this.

Some of what you deliver is knowledge. You know how to do a thing, and people pay you because they don't. Some of it is execution. You reliably do the thing, on time, at quality. And some of it is judgment plus relationship. You know which thing to do, for this specific customer, in this specific situation, and they trust you enough to just say yes.

The knowledge layer is being commoditized and it is not coming back. That's not a prediction, it's already happened, and every month it gets more complete. Anything that could be written down as a procedure is now available to anybody who can describe the problem.

The execution layer is under pressure but it's holding, because reliability turns out to be genuinely hard and mostly not a knowledge problem. Knowing how to do something and doing it forty times a month without dropping any is a different skill entirely.

The judgment and relationship layer is barely touched. Not because machines can't reason, they can. Because that layer depends on context nobody has written down, accountability somebody has to carry, and trust that gets built through a history of showing up.

Most businesses charge for knowledge and deliver judgment. That mismatch used to be fine. It isn't anymore, and closing it is most of what strategy means right now.

Take an hour today. Actual hour, phone in another room.

List every deliverable you sell. For each one, break out roughly what percentage of the value is knowledge, what percentage is execution, and what percentage is judgment and relationship. Be honest, which mostly means being harsher than feels comfortable, because the thing you're most proud of is usually the knowledge part and it's usually the part that's evaporating.

Then ask the question that actually matters. Not "could AI do this," because the answer is increasingly yes for the knowledge portion of everything. The question is: if a competent person with good tools tried to replicate this on their own, what would they still get wrong?

The answers to that question are your business. Everything else is table stakes you happen to currently be charging for.

For my friend with the quoting software, the answer was clear once he stopped panicking. What the weekend rebuild got wrong was every regional code variation, the way a specific supplier's pricing behaves in Q4, the thirty seven edge cases that produce a wrong number and a lost bid, and the fact that when a quote fails at eleven at night before a bid deadline, somebody answers the phone.

The software was never the product. The accumulated correctness was the product, and you cannot prompt your way to four years of finding out what's wrong.

Four things hold up, and only four that I've been able to find.

Proprietary data is the first and strongest. Not data you bought. Data that exists because you operated. What broke, for whom, how often, what fixed it, what the real numbers were versus the quoted ones. Nobody can generate that. They'd have to live it.

Most businesses are sitting on this and not capturing it. Every service call, every project retro, every customer complaint is a data point that becomes a moat if you write it down somewhere structured, and evaporates if you don't. Start writing it down. That's most of the work.

Second is accountability. Somebody has to be responsible when it goes wrong, and that somebody has to have something to lose. This is why regulated and high consequence work stays human longer than the capability curve suggests it should. It isn't about competence. It's about who's liable at three in the morning.

If your work carries real consequences, stop apologizing for your price and start being explicit about what you're carrying. A lot of businesses undersell this because it feels like they're charging for nothing. They're charging for risk transfer, which is one of the oldest and most legitimate things anybody has ever sold.

Third is relationships with actual history. Not a contact list. A history. People who've watched you handle a bad situation well. That has a specific value that nothing replicates, because it's built from time and there's no shortcut through time.

This is where I'd actually spend effort, and where most people spend none. I use Clay to keep track of who I know and when I last talked to them, because the failure mode isn't that relationships end, it's that they quietly go stale while you're busy. A network you actually maintain is worth several multiples of one you merely have.

Fourth is taste. Knowing which of five acceptable answers is the right one here. Impossible to write down, which is exactly why it's durable, and it only comes from having been wrong enough times to develop a feel for it.

Worth noting what did not make that list. Being first does not make it. Working harder does not make it. Having a better version of a tool anyone can buy does not make it. Those were all real advantages once and they're all rentable now, which means they're not advantages, they're expenses your competitors can match on a Tuesday afternoon with a credit card.

The four that hold have one thing in common. Every one of them takes time to build and cannot be bought at any price, which is precisely why they're worth building.

THE AI WORKFLOW BLUEPRINT • $47

Automate the commodity layer so your hours go to the part nobody can copy. Twelve workflows covering the recurring machinery, with the prompts and maps to build them this week rather than someday.

The proprietary data point is the one everybody nods at and nobody acts on, so here's the version that fits in a real week.

You need three habits and none of them take more than ten minutes.

First, a standing question at the end of every project. Not a survey. One line in whatever you already use: what went wrong that we didn't expect, and what did it cost. Write it down even when the answer is nothing, because the nothings are data too and they're what let you spot the pattern when something starts recurring.

Second, capture the near misses. Not just what broke, but what almost broke and got caught. Those are the highest value entries you'll ever write, because they're the edge cases somebody rebuilding your work from scratch would walk straight into. Nobody records these because they feel like non events. They're the entire moat.

Third, tag the reason you lost. Every deal that didn't close gets one of five or six standing reasons, chosen from a fixed list so you can actually count them later. Free text here is useless within a year. Two years of tagged loss reasons tells you more about your positioning than any consultant will.

Store all three somewhere searchable that isn't an individual's head or an individual's laptop. A shared doc is fine. A database is better. The format matters much less than the fact that it exists and that somebody adds to it every week.

The compounding is genuinely slow and genuinely relentless. Six months in it's a curiosity. Three years in it's the thing that lets you quote a job correctly when your competitor is guessing, and quoting correctly is most of what separates the shops that survive a bad year from the ones that don't.

Once you know which parts are defensible, the strategy is straightforward and mostly unpleasant to execute.

Automate the commodity layer aggressively. Not reluctantly. Aggressively. Every hour you spend doing something a tool does at ninety percent quality is an hour stolen from the part of your business that's actually yours. And you're not going to win on price against somebody who automated it, so competing there is choosing to lose slowly.

Reinvest that time entirely into the defensible layer. Capture the data you've been throwing away. Deepen the ten relationships that matter instead of maintaining four hundred that don't. Get better at the judgment calls by doing more of them, and by writing down what you got wrong so the pattern shows up.

Then, and this is the part people skip, change how you talk about what you sell. If your marketing describes the knowledge layer, you're advertising the commodity. Describe the judgment. "We know how to build X" is a claim anybody can make now. "We've built four hundred of these and here's the thing that goes wrong at month six that nobody warns you about" is not.

Reprice while you're at it. If you were charging for knowledge and delivering judgment, you were underpriced and the market just made that visible.

Some businesses don't survive this sort.

If you run the exercise honestly and the answer is that ninety percent of your value is knowledge, with no proprietary data, no accountability weight, no deep relationships and no accumulated taste, then you have a real problem and a genuinely limited window.

The good news is that a limited window is a lot better than no warning. You can build the defensible layer starting now. Start capturing the operational data. Start moving upmarket into work where consequences matter. Start deepening the handful of relationships worth deepening. It takes eighteen months and it's completely doable, but only if you start before the revenue starts sliding, because it's very hard to build a moat while you're firefighting.

My friend with the quoting software raised prices eleven percent and repositioned the whole thing around accumulated correctness rather than features. Lost two customers who were buying on price. Kept everybody who was buying on not losing bids. Revenue's up, headcount's flat, and he sleeps fine now.

Nothing about his product changed. He just figured out what he was actually selling, which turned out to be the thing that was never in the demo.

That's the hour. Go do the sort.

THE AI BUSINESS ACCELERATOR • $97

Eight weeks of building the defensible layer on purpose. We find what's actually yours, automate everything that isn't, and build the systems that capture the data your competitors will never have. Bring the business as it is, not as you wish it were.

Jordan

The AI Newsroom | Jordan Hale | ainewsroomdaily.com