The Talent Pipeline Flush
Every few months, another headline drops: thousands of employees cut at a major tech company, paired with a statement about “repositioning for AI.” The numbers are staggering. The press releases are polished. And if you only read the headlines, you’d think these companies are shrinking.
They’re not. They’re swapping one workforce for another, and the distinction matters.
Microsoft cut roughly 15,000 people in 2025. The same year, it reported $82 billion in annual profit. And when the dust settled, its total headcount was right back where it started—around 228,000 employees.
So what actually happened?
The official explanation was AI. Microsoft needed to restructure around artificial intelligence. Roles were eliminated. The company was leaning into the future. The most significant cuts came in May (~6,000 roles) and July (~9,000 roles).
That explanation is misleading.
What Actually Happened
Microsoft didn’t shrink. It swapped.
Thousands of engineers, product managers, salespeople, and middle managers were let go. And then, quietly, new people were hired into different roles—roles requiring AI and machine learning expertise, roles building out cloud infrastructure, roles with “AI” somewhere in the job title.
IBM did the same thing more openly. It cut around 8,000 jobs in HR and administrative functions while simultaneously hiring engineers and salespeople. Different profile in, different profile out.
This isn’t a workforce reduction. It’s a talent pipeline flush. Companies are cycling out one set of skills and cycling in another. The headcount stays. The humans don’t.
Why the AI Framing Is Misleading
I want to be fair here. AI is doing substantive work at these companies. Satya Nadella said publicly that roughly 30% of Microsoft’s code is now written by AI. That’s a meaningful number. Some of what these engineers were doing is genuinely being automated.
But that’s not the whole story. Not even close.
A significant portion of the roles being cut have nothing to do with AI automation. Middle management layers were flattened. Sales teams were restructured. Gaming divisions were gutted to free up capital. These aren’t jobs AI is doing. These are jobs companies decided they no longer wanted to pay for, in a moment when “AI” provided a convenient and forward-looking way to explain the decision.
Calling it an AI restructuring makes it sound visionary. Calling it a cost-cutting exercise while redirecting capital to infrastructure investment is more accurate—and far less inspiring for a press release.
The data backs this up. A 2026 ResumeBuilder survey of 1,000 hiring managers found that 59% admitted their companies emphasize AI’s role in layoffs because it “plays better” with stakeholders, investors, and the public. Only 9% said AI had actually replaced job roles at their organizations. That’s a 50-point gap between the story companies are telling and what’s actually happening on the ground.
There’s a technical reason that gap exists. AI capabilities don’t arrive in sudden leaps that would justify eliminating thousands of roles in a single quarter. A recent MIT paper on neural scaling laws showed that LLM performance improves along a smooth, predictable curve—loss decreases proportionally to model width, governed by the geometry of how models pack concepts into limited space. The improvement is real but incremental. There’s no inflection point where AI abruptly became capable enough to replace a department. The technology follows a gentle slope, not a cliff. When a company frames a mass layoff as a response to a sudden AI breakthrough, the math doesn’t support that narrative.
The people who lost those jobs built the products that made these companies profitable enough to make an $80 billion AI bet in the first place. They deserve a more honest explanation than “the future requires different skills.”
What This Means for Software Engineers
If you’re a software engineer watching this and wondering where you stand, here’s my honest read.
The threat is genuine, but it’s not evenly distributed. Generalist roles, administrative engineering work, and middle management layers are under pressure. Not always from AI doing the work, but from companies deciding they can do more with less and using AI as justification to make that move.
The demand is also genuine. Engineers who can direct AI tools, build on top of them, and make good decisions about when and how to use them are exactly what companies are hiring for right now. The job postings tell the story clearly: AI-adjacent roles fill in days. Traditional roles sit open for months.
I wrote about this more broadly in “Don’t Panic, Adapt,” looking at the Microsoft research on how AI is reshaping occupations. The takeaway there is the same one that applies here: high exposure doesn’t mean extinction. It means evolution. The real-world usage data shows augmentation dominating over replacement. The people who treat AI as a capable collaborator and focus on the judgment, architecture, and problem decomposition that AI can’t do are the ones landing on the right side of these transitions.
None of this means you should panic. But it does mean you should pay attention.
Don’t Panic, Adapt
At Don’t Panic Labs, we think about this constantly—both for our own team and for the partners we work with.
AI is changing work. That’s true. The nature of software engineering is evolving. Also true. But the engineers I’ve seen struggle through this moment aren’t the ones whose skills became less relevant. They’re the ones who waited too long to engage with what was changing around them.
The engineers who are doing well are curious. They’re picking up AI tools and figuring out what they’re actually good for. They’re finding the places where AI makes them faster and the places where human judgment still matters. They’re not trying to out-compete AI at the things AI does well. They’re figuring out what they can do that AI can’t.
That’s not a guarantee. The transition is happening, and some of the people caught in these layoffs did everything right and still lost their jobs. That’s worth acknowledging—and it’s worth being angry about when companies dress up a pipeline flush as a principled stand for the future.
But for engineers who still have time to act: adapt. Learn the tools. Find where your judgment adds value. Stay curious.
The work is changing. That has always been true in software. The engineers who thrive are the ones who change with it.
