The gap is the window.

A forecast has been making the rounds this year — coming from inside the AI industry itself — that fifty percent of entry-level white-collar work will be “completely wiped out” inside five years. Not automated. Not augmented. Wiped out.

The discourse hasn't really recovered. Most of it has been arguing about whether the number is right. That's the wrong fight. The same firms putting out those forecasts have also been putting out quieter capability research — the work where they actually measure, function by function, what AI can do and what's actually getting done with it. That research tells a more interesting story.

Two numbers keep showing up across it. One measures what AI could in principle do, top to bottom of a business — theoretical capability. The other measures what AI is actually being used to do today — observed reality. The first is large and gets the headlines. The second is small and rarely does. The interesting story is the space between them.

Here's the part that doesn't make the LinkedIn version. The two numbers don't move together evenly. They vary wildly by function.

In a few categories — the work that programmers do, the work that admin staff do — the capability number is high and the observed number is meaningful. Those functions have started to cross. Not by accident. The people inside them were the earliest adopters; they configured the capability to fit their own work because they happened to be exactly the people equipped to do it.

In most other knowledge work — finance, legal, sales, education, content, support — the capability number is just as high. The observed number is near zero. The gap isn't that AI can't do the work. The gap is that nobody's done the work of fitting the capability to the work yet.

In a third group — construction, agriculture, food service, transportation, repair — both numbers are low. The capability hasn't been built; the deployment isn't there. The research says the workforce era looks fundamentally different in those categories than it does at a desk.

The two numbers don't move together. The space between them is where the work is.

The discourse arguing about whether the bigger number is scary is having the wrong argument. The actionable question isn't the macro number. It's the per-function one: in my business, which functions have crossed? Which haven't? Which never will?

Crossing the gap inside your business isn't buying more AI. It isn't writing better prompts. It isn't waiting for a smarter model. The crossing is fitting capability to a specific job — your customers, your team, your prior decisions, the questions your best people actually ask. Without that fit, the same model that could in principle cover your support function gives generic answers on your tickets. With it, it gives the answer you would have given. The work is configuration, not consumption. It looks less like a software purchase and more like an install.

Crossing the gap isn't buying more AI. It's fitting capability to a specific job.

What you get on the other side of the install isn't better software. It's a different scale of business at the same headcount.

A workforce inside a business multiplies what the people there can do. A team of ten starts performing like a team of a hundred or more. Same payroll. Same calendar. The work that used to pile up between meetings — drafted, qualified, researched, followed up, escalated, closed out — gets done by the time anyone sits down to do it.

The operators who haven't crossed are running the same business at the same speed they always were. Their peers stopped competing with them six months ago.

A team of ten performs like a team of a hundred — same payroll, same calendar.

The forecast that fifty percent of entry-level work would disappear is, technically, a forecast about labor markets. Read against the actual capability research, it's something different. It's a description of what happens when a function goes from “in principle” to “in production.” The work doesn't vanish. The shape of who does it changes. Some businesses will close their gaps with smaller teams. Some will close their gaps with the same team running ten times the surface area. The forecast picked the most dramatic version. The research's actual finding is that the choice is yours — and it's per function, not industry-wide.

The forecast was loud. The actual finding is that the choice is yours.

Two numbers keep showing up. One is what AI could do across every function in your business. The other is what AI is doing today. The space between them is the operating window. Every business has one. Windows close.

The gap is closing. It's closing for someone.