The Companies Building AI Just Proved Engineers Aren't Going Anywhere

SignalFire tracks hiring data across 80 million companies. Overall tech hiring is still stuck at 75% of pre-pandemic levels. Engineering headcount? Barely moved. This week the Wall Street Journal profiled OpenAI, Google, and Anthropic—the three companies building the AI that's supposedly automating engineers out of existence. They found AI running through every process, deployed in every workflow. And the engineers? Still very much in the building.

I've been through four of these cycles now. The spreadsheet was going to eliminate finance jobs. The internet was going to disintermediate every professional service. Automation was going to hollow out manufacturing management. Each time, the prediction was the same: technology eats the humans. And each time, what actually happened was more interesting—and more uncomfortable.

The Layoff Story Has It Backwards

Here's the pattern everyone missed: AI didn't kill the work. It moved the bottleneck.

When the model writes the code in seconds, execution stops being the constraint. When it generates ten versions of the analysis before lunch, production isn't your problem anymore. What's scarce now is judgment—knowing which of those ten outputs is actually any good. And orchestration—pointing people and machines at the same outcome without them working against each other.

I was working with a financial services client last month. They deployed Copilot across their development team, expecting to cut contractor hours. Three months in, their senior engineers were working longer, not shorter. Why? Because the machine was generating more code than the organization could evaluate. Someone still had to decide which implementation was production-ready, which shortcut would create technical debt, which "solution" would fail under load. The machine is extraordinary at generation. It's useless at knowing which answer actually mattered.

Nobody Gets Fired For Typing Slowly

The spreadsheet is the cleanest parallel here. In 1979, VisiCalc shipped and the predictions were immediate: finance jobs would disappear. Why pay an accountant to do math when the computer does it instantly?

What actually happened? Spreadsheets didn't thin out finance departments. They raised the price of the people who could read the numbers and decide what they meant. The machine handled calculation. It made the humans who understood business context, risk, and strategic implication more valuable, not less. The analyst who could look at a model and say "your assumption in cell D47 is wrong and here's why" became irreplaceable.

The scarce resource was never typing speed. It was knowing what good looks like.

The Rick Rubin Problem

Rick Rubin can't play an instrument. He can't run the board. His value is taste—he knows what "done" sounds like before anyone else in the room. Phil Jackson never outscored one of his players. His value was orchestration: he built the conditions where that much talent could actually win championships instead of flaming out in the second round.

This is where AI is taking us. The question isn't "can the machine do the task?" It's "who knows when the output is right?" The machine generates options faster than any human ever could. If you're the person who looks at what your team—human or machine—produced and says "not that, this," you didn't get less valuable. You got scarcer.

I'm watching engineering teams struggle with this right now. Junior developers used to learn by writing basic CRUD operations, fixing bugs, grinding through the repetitive work that built pattern recognition. AI does that now. So how do you develop judgment when the machine handles everything judgment is built on? That's the uncomfortable question nobody's answering yet.

What Your Hiring Data Should Be Telling You

Go look at your own numbers. If you're a finance leader, an audit partner, anyone running a team that uses AI tools—where are you actually cutting? My bet: you're trimming execution roles and desperately hunting for people with judgment. The person who can review an AI-generated audit procedure and spot what it missed. The analyst who knows when the model's recommendation is technically correct but strategically stupid.

This creates an ugly dynamic that nobody wants to say out loud. The career ladder just lost most of its rungs. The path used to be: do grunt work, build skills, develop judgment, become valuable. Now the grunt work is gone. How do you build judgment without repetition? How do you evaluate whether someone has taste before they've shipped enough to demonstrate it?

The Question That Ages Badly

So if you run a team or a budget, here's the decision in front of you: are you modeling AI as a way to delete headcount, or as the thing that makes judgment your most important hire?

One of those plans ages badly.

The executives who treated spreadsheets as a headcount reduction tool spent the 1980s wondering why their competitors were making better decisions faster. The ones who realized spreadsheets made analysts more powerful built finance teams that became strategic partners instead of cost centers.

I can't tell you exactly what the org chart looks like in three years. But I can tell you what I'm seeing in the companies that aren't panicking: they're hiring differently. Less "can you code?" More "can you tell when code is wrong?" Less "can you run the analysis?" More "can you spot when the analysis is answering the wrong question?"

The machine made judgment expensive. Plan accordingly.

What To Do Monday Morning

Here's your checklist:

Ask your team leads: When we deploy AI tools, what are people spending their freed-up time on? If the answer is "more AI-generated output," you're building a volume problem, not solving one.

Look at your hiring rubric: Are you screening for execution speed or decision quality? If your interview process still optimizes for "can they do the task," you're selecting for the skills the machine just commoditized.

Audit your development pipeline: How are junior people building judgment when the machine handles the work that used to build judgment? If you don't have an answer, you're going to have a senior talent crisis in 24 months.

The scarce resource isn't changing. It was always judgment. AI just made it more obvious.

What are you hiring for?