Why companies are cutting roles in the AI era, and what to do about it

When a large company announces layoffs and points at AI, two very different things can be happening, and they look identical from the outside. For anyone thinking about AI and small business staffing, the difference is worth reading carefully, because the headlines shape decisions you might copy without meaning to. Sometimes the company genuinely found that software now does work people used to do, so it needs fewer people. Sometimes the company wanted to cut costs anyway, for reasons that have nothing to do with AI, and "AI made us more efficient" is a cleaner press release than "we over-hired and the market turned." Telling those two apart matters, because only one of them tells you anything useful about your own job.

Real restructuring versus a convenient cover story

Real, productivity-driven restructuring has a signature. The work that used to take a team genuinely gets done by fewer people plus software, and output holds or rises. You can see it in what still ships after the cut. The roles that disappear are the ones where the task was repetitive and well-defined enough that a tool could absorb it.

The cover-story version looks different underneath. The company was going to trim regardless, the business reasons are ordinary, and AI is the framing that makes Wall Street nod and makes the cut sound like strategy instead of retreat. Saying "we are leaner because of AI" sounds forward-looking. Saying "demand softened and we hired too fast" sounds like a mistake. Both cuts get the AI label, but only the first one reflects work actually being automated.

For your own planning, do not take the announced reason at face value, in either direction. Watch whether the work genuinely got absorbed by tools or whether the company simply needed to spend less and reached for the explanation that played best. The label tells you about the company's messaging. The behavior of the work tells you the truth.

What actually gets automated first

Strip away the framing and a consistent pattern remains. The tasks most exposed are the ones that are repetitive, well-specified, and judgment-light. Work that follows a clear procedure, produces a predictable output, and rarely surprises anyone is the work software does well. If your day is mostly steps a tool could follow without you, that is the part under real pressure, regardless of which company is in the headlines or what reason they gave.

That is not a doom sentence. It is a map. It tells you exactly which parts of your work to stop relying on as your value, and which parts to lean into.

Make your work hard to automate

The durable response is to deliberately move your value toward what tools are bad at. A few things stay hard to automate for a long time.

Judgment in messy, ambiguous situations, where the right answer depends on context no model has and the cost of being wrong is real. Owning an outcome rather than executing a task, where someone has to decide what should happen and be accountable when it does not. Work that needs trust and relationship, where people are buying your judgment and your reliability as much as the deliverable. And the ability to use the tools well, directing them, checking their output, and catching the confident mistakes, which is itself a skill that rises in value as the tools spread.

The move is not to compete with the tool at the thing the tool is good at. You will lose that, and the loss gets worse every year. The move is to become the person who aims the tool, judges its output, and owns the result. Take the tasks that are getting automated and let them go. Pour the time you get back into the parts of your work that need a human who can be trusted with an ambiguous, high-stakes call. That is the work that is hard to cut no matter what reason ends up in the press release, because it was never the kind of work software was going to do.

Related reading

- [Stop chatting with AI and start handing it real work](02-ai-delegation-loop.md)

- [Treat AI output as a first draft, never a finished product](08-ai-output-first-draft.md)

- [What it means when a company treats AI like a team member, not a tool](13-ai-as-team-member.md)