"Your competitors just bought the same AI as you. Now what?"

Last week a post did the rounds on LinkedIn claiming the moat in AI isn't any single tool — it's stacking a handful of them together into a system that runs itself. A hundred-odd replies later, something unusual had happened: everyone agreed with the premise. Tools aren't the moat. The entire fight was over what is.

Three camps formed. One said the moat is architecture — tools get swapped out, but a well-designed operating model survives the swapping. Another said it's engineering discipline — one commenter pointed to a real security hole in AI-generated apps that exposed the data of more than 170 businesses, and asked the question demos never answer: who's awake at 2am when the automation chain snaps? The third camp said the tech was never the bottleneck at all. AI made building cheap. It didn't make customers cheap, and it didn't make a business ready to absorb what the tools produce.

Here's what struck me reading it. Every person in that thread is a builder, arguing about builders' moats. And every moat they proposed can be bought. Architecture is hireable — any competent firm will design you a durable one, and the same firm will design one for your competitor next quarter. Engineering discipline is hireable too; that's exactly what good contractors sell. Even distribution can be rented for a while, if you're prepared to pay for it. The whole debate was people with the same skills arguing over which of their skills matters most — and missing that a skill anyone can hire is not a moat for anyone.

Now bring it back to a business like yours — the family operation running showroom, factory, field and install under one roof. You've been watching this AI wave with a fair question: if all this capability is suddenly cheap, and the roll-up down the road can buy the same tools we can, where does that leave us?

It leaves you holding the one input that entire thread forgot to mention.

Every system those builders were arguing about is an empty machine until something goes into it. The stack is a commodity — they're selling it to everyone, that's the business model. The architecture is a service. The discipline is a day rate. What's none of those things is the twenty-odd years of judgment that prices your jobs, reads your builders, knows which supplier's lead time to believe and which customer is worth sharpening the pencil for. That's the input. Nobody in that comment thread can sell it, hire it, or generate it — because it doesn't exist anywhere except inside your operation.

The filter, if you want one: if a competitor can buy it, it's not your moat. Run that over everything an AI vendor shows you this year.

The tool — buyable by definition; it's for sale, that's why they're in the room. The build — buyable; someone built it for you, someone can build it for them. The judgment and the data that go in — that's the first thing on the list no competitor can purchase at any price. Which means the strategic question was never "which AI tool should we buy." It's "what do we know that nobody can subscribe to — and have we captured it anywhere?"

One thing that thread did get right, and it's worth taking seriously. The engineering camp's warning is real. AI-built systems break — quietly, in ways the demo never shows — and somebody has to be responsible when they do. For an operation like yours, that's not a reason to sit this out. It's the reason the owner's sign-off stays in the loop. A system built from your judgment, with you approving what it puts out until it's earned your trust, fails safe. A cheap automation nobody in the building understands fails at 2am.

The builders will keep arguing about the machine. They're good at machines — it's their trade. But the machine was never the scarce part. What's scarce is what you'd feed it. And right now, in most family operations, that's written down nowhere, held by one or two people, and walking out the door the day they do.

I've built a short assessment for multi-channel family operations working out where AI actually helps — and which part of their edge is worth capturing first. Ten questions. The readout comes from me, not a machine.

Your business was too complex for standard software: why that's now your advantage …

You know the moment. You're in a demo, or on a call with a software rep, and you're explaining how the business actually works — the showroom out front, the factory out back, the install crews, the trucks, the fact that a job doesn't move in a straight line because real jobs never do. And somewhere in there the rep's face changes. They nod, they say "right, right," and then they steer you back to the part their product handles. The rest — the part that's most of your business — gets folded into "we can look at that in a later phase."

What just happened is the whole story. You described your business, and the tool asked you to become a simpler version of it so the software could cope.

Most owners walk out of that room feeling like the problem is them. That the business grew up messy, that it should have been built cleaner, that everyone else must have this figured out. I want to take that off you, because I spent ten years on the other side of that table — selling the software — and I can tell you exactly what was going on. It wasn't your business. It was the model the product is built on.

Software scales by serving the average. That's not a criticism, it's just the economics: a product makes money by finding the biggest group of customers who all have roughly the same problem, and building the one thing that solves it for all of them. The more customers who fit the same mould, the better the business. Which means every software company is, by design, hunting for the average — the median business, the standard workflow, the job that moves in a straight line. That's who the product is for.

The family operation running retail, manufacturing, logistics and install under one roof is the exact opposite of that. You're not the average of anything. Your business is a one-off, built over a generation, shaped by decisions nobody else made in the same order. The very thing that makes you you — the way the showroom feeds the factory feeds the install crew, the judgment that ties it all together — is the thing that fits no mould. So the software can't hold you. Not because it's bad software. Because you are, precisely, the customer the model is built to exclude.

For twenty years that felt like a curse. You watched simpler businesses buy tools that fit them out of the box, while you ran on spreadsheets, memory, and the owner's head. Now the picture flips.

Generic AI just got cheap. Everyone can buy it — the roll-up, the franchise, the operator down the road. And what generic AI is good at is exactly the average stuff: the standard task, the common workflow, the job that moves in a straight line. It commoditises the median beautifully. Which means the median is now worth nothing as an edge, because everyone has it at the same price.

The one thing it can't commoditise is the thing that was never average to begin with. Your complexity. The cross-functional judgment that reads a job across make, sell, move and install all at once — that no product could ever hold — is now the last defensible edge in your business. The moat isn't despite the complexity. It is the complexity.

Here's the insight worth taking away, because it's where most people get this wrong. There's a new wave of AI-native tools built specifically for trades — glass, stone, joinery, fabrication — and they're good. Genuinely. But watch where they stop. They handle the rule-like part of the business with no trouble: standard pricing, supplier rates, the jobs that follow a formula. That's the layer where the same inputs always give the same answer, and a tool eats it easily. Where every one of them stops dead is the cross-functional read — the price that depends on the factory being flat out that week, the access being tight on that site, the builder being someone you'll sharpen the pencil for, and the last three jobs for them running late. That judgment doesn't live in any one function, so no single-function product can reach it. It's not a gap they'll close in the next version. It's structural. The tool is built to serve one department; your edge lives in the seams between all of them.

So the filter, when the next AI tool lands in your inbox: does this handle my rule-like work, or is it claiming to handle my judgment? Buy the first kind cheaply and don't overthink it — fighting a good tool over standard pricing is wasted money. Be very careful with anything claiming the second, because the judgment is the part that's yours, and the part worth owning rather than renting.

The businesses that spent twenty years being told they were too complicated to serve were right to protect what made them different. That complexity was never the weakness. It was the edge — waiting for the moment the average became worthless. That moment is now.

I've built a short assessment for multi-channel family operations working out where AI actually helps — and which part of their edge is worth capturing first. Ten questions. The readout comes from me, not a machine.

What Big Enterprise is already doing with AI (what it means for you)

I spent ten years selling software into big companies. Which means I spent ten years in the offices where the enterprise version of whatever's coming next gets sold first — the boardroom pitch, the pilot, the rollout, the post-mortem eighteen months later when it didn't do what the slide said.

The useful thing about having sat on the vendor's side of that table is that you learn to read the play early. What the big end is doing this year, the mid-market gets sold next year, usually worse and more expensive. So it's worth knowing what's actually happening, not the headline, but the reality, because it tells you what's heading for you and what to do before it arrives.

Here's the honest version, stripped of the noise.

The big companies have moved past playing with AI. Two years ago it was experiments — a chatbot here, someone in marketing trying things. Now it's in production, running real work: agents that read the documents, pull from the systems, apply the rules, and actually do the task rather than just answer a question about it. Salesforce, Microsoft, ServiceNow — the platforms your bigger competitors buy from — have rebuilt their whole model around it. Microsoft alone has hundreds of thousands of these agents running across companies right now. This isn't coming. It's here, at the top end, today.

Now the part nobody puts on the slide. Most of these projects are struggling, and not for the reason you'd think. They're not failing because the AI isn't clever enough — the AI is plenty clever. They're failing on three things, and I want you to read these carefully, because they're the whole point of this piece:

They're failing because the data was a mess. The AI is only as good as what you feed it, and most big companies discovered their information was scattered, inconsistent, and half-wrong the moment they tried to build on it.

They're failing because nobody owned the edge cases. The agent handles the standard job fine, then hits the one-in-five job that's unusual, and there's no clear answer to "who decides what happens now?" So it guesses, and gets it wrong, and someone finds out later.

And they're failing because nobody built the part where a human checks the work before it goes out. They automated first and asked "who's responsible when this is wrong?" second — which is the wrong order, and an expensive one.

The companies getting it right are doing the opposite of the big splashy rollout. They pick one job — one well-defined job with good, clean information behind it — put the AI on that, measure whether it actually worked, and only then do the next one. Narrow, proven, expanded. Not broad, hopeful, and prayed over.

Read those failures and that fix back, and here's what should jump out at you: the big end is spending millions of dollars and eighteen months each learning a lesson you already know in your bones.

You already know your data is messy — it's in your head and a dozen spreadsheets, but you know where the bodies are. You already know the edge cases, because you're the one who handles them; the unusual job doesn't get guessed at in your business, it comes to you. And you already know a number shouldn't go out the door without someone who knows what they're doing signing off on it — that's just how you've always run. The three things sinking million-dollar enterprise projects are three things a good family operator does by instinct.

That's the whole advantage, and it's worth saying plainly. The big end has scale, budget, and a stack of consultants, and it's still tripping over data, edge cases and accountability. You have less budget and no consultants, but you have the three things that actually matter already sitting in the business: you know your work, you own your judgment, and you're already the human who checks it. The enterprise play, done properly, is just: start narrow, on a job you understand, with you signing off until you trust it. You don't need a million dollars to run that play. You need to run it before the watered-down, oversold version of it turns up in your inbox with a logo on it — which it will.

So the one move, before that happens: pick the single job in your business where the same kind of decision gets made over and over, where you've got real history behind it, and where you'd sleep fine knowing a system drafted it as long as you still signed it off. That's your one narrow job. That's where this starts — not with a platform, not with a big rebuild, and not with whatever the enterprise vendors will eventually try to sell you. With one job you already understand better than any software company ever will.

I've been on both sides of this — the enterprise floor where the play gets sold, and the operations floor where the work actually happens. The gap between those two rooms is where most of the money gets wasted. It doesn't have to be wasted at your end.

I've built a short assessment for multi-channel family operations working out where AI actually helps — and which part of their edge is worth capturing first. Ten questions. The readout comes from me, not a machine.

Your DNA.The most valuable thing in your business.

Every family business (or any small business) I've ever been inside runs on the same thing, and it's never in the files. It's the way one or two people read a job. What it'll really cost. Where it'll go wrong. Which customer is worth sharpening the pencil for and which one to let walk. Which supplier's word to trust and which to double-check. That reading — built up over twenty or thirty years, mostly without anyone noticing it was being built — is what actually holds the business together. Not the software, not the systems, not the price list. The judgment.

I grew up in one of these. My family had a trade business, and I did the sign-writing for others like it before I ran my own small printing trade. So I know how this knowledge lives. It doesn't live on paper. It lives in a person, and it comes out in the moment — a glance at a job, a number that just feels right, a "no, not that one" that turns out to be spot on for reasons nobody bothered to spell out. It's real, it's valuable, and it's the least written-down thing in the whole operation.

Here's the quiet problem with that. The most valuable asset in the business is also the one nobody else can touch. It can't be shared, because it was never put into words. It can't be built on, because there's nothing to build on — it's in one head. It can't be owned by the business, because it's owned by a person. And none of that is a problem on any given Tuesday. It only becomes a problem the day that person wants a proper holiday, or wants to hand more across, or just wants the business to be worth something that doesn't depend entirely on them being in the room.

Most families never deal with this, and I understand exactly why. The only version of the conversation anyone knows how to have is the awkward one — the "so, what's the plan, when are you thinking of stepping back" one. Nobody wants to start that. It sounds like you're measuring someone for the exit. So it gets left, year after year, until something forces it, and then it's rushed and tense and done badly.

I want to offer a different conversation, because there is one.

It isn't about anyone stepping back. It's about capturing what one person knows while they're right there, engaged, and at the top of their game — turning the judgment that's run the business into something the business actually owns. Done properly, this isn't a founder being downloaded before they're pushed out the door. It's the opposite. It's treating that person's know-how as the skilled, hard-won thing it is, and finally giving it the respect of writing it down properly — so it can be taught, trusted, built on, and passed on, with them leading it rather than being sidelined by it.

And here's the part most people haven't considered, because they picture the wrong thing. When they imagine "capturing the knowledge," they picture a fat questionnaire, or a consultant with a clipboard asking someone to explain their gut feel — which never works, because you can't explain gut feel to order. That's not how it's done. The way it actually works is you sit with the person while they do the real thing. You watch them price an actual job, and you get them talking it through as they go — the asides, the "well, normally I'd, but this builder always," the little adjustments they make without thinking. That's where the judgment lives, in the asides. You don't interrogate it. You watch it work, and you catch it.

Do that well and something unexpected happens. The person whose knowledge it is tends to enjoy it — because for the first time someone's treating the thing they're best at as worth understanding properly, rather than as an obstacle to a system. It turns out to be one of the better conversations a family business can have. It brings people around the thing they're all proud of, instead of the thing they're all sometimes avoiding.

The judgment in your business is the most valuable thing you've got, and right now it lives in exactly one or two places and nowhere else. That's not urgent today. It's just true. The best time to capture something like that is long before it's urgent — while the person who has it is there to lead it, and it can be done properly and without pressure, as the respectful thing it ought to be.

I've built a short assessment for multi-channel family operations working out where AI actually helps — and which part of their edge is worth capturing first. Ten questions. The readout comes from me, not a machine.