Your Job Is Next

A few days before I sat down to write this, a company laid off 30,000 employees in a single move — not because the business was struggling, but because AI could do the same work cheaper, faster, and without needing a break. That company wasn't an outlier. Amazon cut a comparable number across two phases. Microsoft, Meta, Intel, Nissan — the list of companies making the same calculation keeps growing.

This isn't a prediction about the future. It's an inventory of the present. The question was never whether this wave would arrive — it's where you happen to be standing when it does. The World Economic Forum's 2025 report puts a number on it: roughly 92 million jobs affected in just the next five years. That's not a script from a science-fiction film. That's a research institution's read on the next five years, full stop.

The jobs we assumed were safe

For a long time, the working assumption was that repetitive, manual roles — factory work, data entry — were the exposed ones, while anything requiring judgment or specialized training was insulated. Lawyers, doctors, accountants, software engineers: all reasonably confident their work was too nuanced to automate.

Turns out, confidence wasn't evidence. JPMorgan's AI system, COIN, reviews 12,000 loan agreements in the time it takes to read this sentence — work that used to consume 360,000 hours of legal team time a year. The company saves an estimated $50 million annually from that single deployment, with no fatigue, no errors from a long shift, no time off.

Medicine tells a similar story, but with higher stakes. Joseph Coates was diagnosed with POEMS syndrome, an exceptionally rare blood disorder. His doctors gave him nothing — a choice between dying at home or dying in a hospital. An AI system found a medicine combination human specialists hadn't identified, and it worked. Joseph is alive. And because that discovery is now documented, every future patient with the same diagnosis inherits it.

Even the people building these systems aren't exempt. AI coding assistants now write an estimated 46% of code with no human involvement at all. The engineers constructing the thing that's automating other people's jobs are watching it start to automate parts of their own.

The danger was never in the jobs AI can replace — it's in never asking the question and staying at the bottom layer.

Three layers of work

Here's a framework I've found useful for cutting through the noise, and I'd encourage you to run your own job through it honestly.

Work splits roughly into three layers. At the bottom is skilled but repetitive labor — tasks measurable against a fixed set of rules: data entry, first-line customer support, boilerplate coding. This layer is being automated fastest, and it isn't close.

In the middle is management in the fuller sense — not just "team lead" as a title, but anyone navigating complex, human-facing situations and making judgment calls under ambiguity. AI can finish discrete tasks here, but it isn't ready to guide human relationships. Not yet, anyway.

At the top are the innovators — the people identifying problems nobody's named yet and building the systems to solve them. This layer isn't threatened by AI in the near term. If anything, AI becomes a force multiplier here, letting people build faster than they ever could alone.

The exercise worth doing, once you've read this far: set aside your salary and your job title, and actually look at what you spend your working hours on. Which layer does most of it fall into? Are you being pushed toward the top layer, or are you standing still in the bottom one, hoping the question doesn't come looking for you?

Because it will. The risk was never really in the jobs AI can do. It's in not asking this question at all.