AI Is a Comprehension Transfer
A door in my house needed a $2 shim. I'd priced the replacement at $1,000. The model didn't fix it, it explained it, and that distinction decides which of your revenue survives.
A door in my house wouldn't shut. It had been getting worse for months, catching against the jamb, and I'd reached the point of accepting that this was a job for somebody else.
I'd already priced it. Replacing the door and the frame was going to run $1,000, probably more.
It needed a $2 shim behind the bottom hinge.
I'd been trying to solve it from the top, on the theory that a door dropping at the latch side needs lifting at the top hinge. That's wrong, and it's wrong in a way that feels right. You shim the bottom hinge, which pushes the door up. I worked through it with a general-purpose model over six or seven exchanges, most of them useless, and it surfaced somewhere near the point where I'd decided to give up.
Here's what matters about that, and it took me a while to see it.
The model didn't fix the door. It didn't supply hands, a screwdriver, a shim, or a Saturday afternoon. I already had all of those. Every physical input required to solve the problem was sitting in my house the entire time.
What it supplied was understanding.
The distinction that matters
We keep talking about AI as a capability transfer. It can do things now that only people could do, so the people who did those things are exposed.
That's not what happened to my door, and I don't think it's what's happening generally.
AI is a comprehension transfer. It moves understanding from the people who have it to the people who don't, at zero marginal cost, instantly, at eleven at night, without an appointment.
Which produces a much sharper claim than the usual one:
Every dollar you earn from explanation is at risk. Every dollar you earn from execution is not.
Those are different businesses that frequently share an invoice, and most people have never separated them, because until recently there was no reason to.
The question underneath it
Before the framework, the question the framework exists to answer:
What would your customer stop paying you for if they understood it as well as you do?
It's uncomfortable, it's answerable, and unlike most strategy questions it produces a number. Most people can get within 20 percent of it in an afternoon. Very few have tried.
Two axes, four positions
Ask two things about any customer. Do they understand their problem, and could they execute the fix.
| | Can't execute | Can execute | |---|---|---| | Doesn't understand | Dependent. Safe today. | Exposed. The wasting asset. | | Understands | Durable. Real value. | Gone. Already left. |
Here's the mechanism, and it's the whole thing.
AI only moves customers rightward. It does not move them up.
Licensure doesn't shift because a model got better. Neither does liability, specialized equipment, physical presence, time, or scale. Nobody with water coming through their ceiling is iterating with a chatbot. What shifts is comprehension, and comprehension is the horizontal axis.
So two migrations happen at once, and only one of them is a loss.
Exposed to Gone. The trip charge for a three-minute fix. The replacement that wasn't necessary. The audit that was a checklist somebody could have run themselves. That revenue is leaving and it isn't coming back, and it's worth being honest about what it was: rent on an information gap.
Dependent to Durable. This one isn't a loss, it's a repricing. The customer now understands what you're doing and why. You lose the margin you were getting from mystique. You gain a customer who can tell the difference between you and the operator who was going to sell them a $1,000 door.
The part nobody is writing
Honest operators net out ahead on this.
The revenue AI is destroying is disproportionately the low-integrity revenue. Unnecessary replacements. Inflated diagnoses. The charge you resented paying and paid anyway because you couldn't evaluate it. If your business depends on the customer not knowing what you know, that's a genuine problem and it's arriving.
If your business depends on skill, licensure, liability, and being physically present with tools, you're fine. Better than fine, because you're now much easier to distinguish from the alternative.
This isn't a disruption story. It's a balance-sheet reclassification. Some of what got booked as revenue was rent on an information gap, and the gap is closing. The line item was always mislabeled. We're just finding out now.
Two things I'm not claiming, because the strong version of this argument is wrong.
I'm not claiming AI is killing the trades. Licensed, liability-bearing, emergency, tool-dependent work is untouched, and will stay untouched.
And I'm not extrapolating from myself. I'll iterate through six wrong answers on a hinge and I own tools. Most people won't and don't. Realistically this moves jobs-a-homeowner-will-actually-attempt from something like a fifth to maybe a third. That's real and it's meaningful and it is not an apocalypse.
Where this lands hardest
Now the part that isn't about doors.
A large share of what gets sold as professional services is comprehension revenue.
The quarterly review that explains a customer's own usage data back to them. The strategic guidance that is, honestly, a recitation of best practices. The onboarding that walks somebody through documentation that already exists. The training module that transfers a concept.
Every one of those is explanation. And explanation is now free, instant, and available at eleven at night without a calendar invite.
I'd include a lot of what currently sits inside customer success in that. Not all of it. But the parts that consist of a person on a call transferring understanding to another person are competing, right now, with something that does that at zero marginal cost and never has a scheduling conflict.
What survives is intervention.
Knowing which account is about to hit an unmet need, and doing something about it at the moment it matters. That's execution. It sits in Durable. And it gets more valuable as the explanation layer commoditizes, because it becomes the only part a customer can't self-serve.
The sentence I'd write on the wall
Here's the consequence that I think has not landed yet with anyone.
You can no longer blame slow value realization on customer comprehension.
That excuse worked for twenty years, and it worked because it was often true. Customers were slow because they didn't understand, and helping them understand was billable. Services, training, onboarding fees, an entire category built on a bottleneck.
It isn't the bottleneck anymore. Your customer can get the concept explained to them, competently, in about ninety seconds, by something that doesn't need a purchase order.
So when value realization is slow now, that's a design failure. Yours. It means the thing is hard to use, or the sequence is wrong, or nobody told them what to do at the moment they were ready to do it.
That's a harder problem than explanation. It's also the one you actually get paid for.
There's a second question underneath this one, which is what's left of expertise at all once explanation is free. The short answer is that what survives is the sequence, not the answer.