For decades, a company's org chart was also its theory of work. Growth, Revenue, Product, Operations, each box a function, each function a place where a certain kind of task lived. If you needed something done, you knew which department to walk into.
That theory is breaking, not because the work disappeared, but because the boxes were never really about the work. They were a proxy for something else: where judgment happened to concentrate.
What's actually being compressed
Look inside any of those departments right now and the routine layer is thinning fast. Outbound sales outreach that used to occupy a team of SDRs runs largely on automation. Junior content that used to take a writer a day now drafts in minutes. Operations workflows that used to require a coordinator checking boxes now run on agents. Entry-level design work that used to be a Photoshop apprenticeship is increasingly a prompt.
None of this means the departments are empty. It means what's left inside them has changed shape. PwC's 2026 Global AI Jobs Barometer, drawn from more than a billion job postings, found that jobs explicitly requiring AI skills are growing roughly eight times faster than the job market overall, with a wage premium that's climbed to 62%. More strikingly, in AI-exposed roles, entry-level postings are now seven times more likely to require the kind of judgment and leadership that used to be reserved for people years into a career.
That's the tell. The market is not simply asking for fewer people. It's asking for a different kind of contribution from the people it retains, the thing the automation can't do: make a call when the answer isn't already written down.
AI handles what can be specified. Humans handle what requires judgment.
That's the actual dividing line, and it cuts straight through every one of the old department boxes rather than sitting neatly inside any one of them.
A task that can be fully specified, here's the input, here's the output, here's the rule for getting from one to the other, is a task AI is already good at, and getting better at quickly. A task that requires judgment, deciding what "good" even means in a specific context, weighing a trade-off nobody wrote a rule for, catching that something is subtly wrong before it becomes a real problem, is a different kind of work entirely. It doesn't compress the same way, because there's no procedure to hand to a model in the first place.
This is why the old function names are starting to feel like they're describing the wrong thing. "Growth" was never really one skill, it was outreach that could be systematized sitting next to relationship judgment that couldn't. "Ops" was never one skill either, it was repeatable process sitting next to the judgment of knowing when the process itself needs to change. AI is peeling those two things apart inside every department at once. What's left behind isn't a department. It's a domain of judgment that used to hide inside one.
Where the new domains are actually showing up
They don't look like the old org chart, and they don't yet have the tidy names departments used to have. They look like: knowing how to actually get AI working inside a real business, not the pilot, the whole workflow, with the change management that goes with it. Knowing how to evaluate whether an AI's output can be trusted, and building the discipline to catch it when it can't. Knowing how to translate a regulation or a governance framework into something a team can actually operate by, not just a document nobody reads. Knowing enough inside a specialized field, medicine, law, a local language, an industry's specific practice, to be the human ground truth a model has to be checked against.
None of these existed as job categories five years ago in the form they're taking now. All of them require the same underlying thing: someone whose judgment can be relied on, not just someone who can execute a step. Trust isn't a separate quality alongside judgment here. It's what judgment earns once it's been demonstrated enough times to be relied on without checking.
The org chart will keep changing. The judgment won't move as fast.
Companies rebuilding their teams around this shift face a genuine design question, and it's not "which department do we cut." It's "where does judgment actually need to sit now, and who's ready to hold it." That's a harder question than headcount planning used to be, because the answer doesn't map cleanly onto Growth, Revenue, Product, and Ops anymore.
The next generation of companies may not be organized primarily around functions at all. They may be organized around where judgment is needed, where it can be safely handed to AI, and where it still can't.
Source: PwC, 2026 Global AI Jobs Barometer.