Amazon just announced 14,000 layoffs, its largest ever. Microsoft has cut around 15,000 this year. IBM, CrowdStrike, and dozens of others have followed. The common thread in every press release and CEO memo: AI.
“AI is the most transformative technology since the Internet.” “We need to be organized more leanly.” “We’re reimagining our mission for a new era.”
I’ve read these memos. The language is always the same. And the logic is always backwards.
The Task, Not the Person
AI replaces tasks. Not people.
This distinction sounds pedantic but it’s the entire ballgame. Any knowledge worker’s job involves dozens of distinct tasks: gathering data, cleaning it, running calculations, building presentations, writing documentation, reviewing requirements, communicating with stakeholders. AI might automate some of those tasks beautifully. But the person who understands the context, who knows which questions to ask, who can spot when the AI’s output doesn’t pass the smell test? That person is more valuable with AI than they were without it.
When you fire the person and keep the AI, you lose the judgment layer. And the judgment layer is the thing that makes AI outputs useful instead of dangerous.
What I’ve Learned Across Four CTOs’ Worth of Experience
I’ve led engineering organizations at Zipcar, Flexcar, and now Vestmark. I’ve built teams at Amazon and Wayfair. I was Chief Architect at Drift when conversational AI was first becoming a product category. Across all of those contexts, the lesson has been the same: technology transitions reward the organizations that invest in their people, and punish the ones that treat people as interchangeable with their tools.
At Zipcar, we automated a huge number of fleet management tasks that used to be manual. We didn’t fire the fleet operations team. We gave them better tools and pointed them at harder problems. The automated systems needed human oversight, edge case handling, and continuous improvement. The operations team became more valuable because they could focus on the work that actually required human judgment.
Here’s what I do when AI changes the work:
Redeploy. When AI automates a task, the person who used to do that task moves to higher-value work. The engineer who used to spend two days writing boilerplate code now spends those two days on architecture and design. The total output goes up. The headcount stays the same. The value per person increases.
Upskill. Every person in the organization needs to understand how to work with AI effectively. That’s a survival skill. We invest in training, in internal tooling, and in creating an environment where people can experiment with AI without fear that they’re automating themselves out of a job.
Learn. This is the part the layoff-first companies get catastrophically wrong. When you’re deploying AI into production, the learning curve is steep and the feedback loops are long. You need people who understand the domain deeply enough to evaluate whether the AI is helping or hallucinating. Fire those people and you’ve cut the feedback mechanism that makes your AI deployment successful.
The Strategy Tax
I call the layoff-first approach a “strategy tax” because it feels like a strategy but actually makes your real strategy harder to execute.
You save money in Q4. You report it to the board as an efficiency gain. The stock price ticks up. And then in Q1, you discover that the people you laid off were the only ones who understood the legacy system your AI needs to integrate with, or the compliance framework your AI needs to operate within, or the customer relationships that inform what your AI should actually be doing.
Now you’re hiring contractors at twice the cost, or shipping AI features that miss critical requirements, or building products that solve problems your customers don’t have.
The strategy tax is real and it compounds.
A Different Playbook
If you’re a leader thinking about how AI should change your organization:
Map your tasks, not your roles. Understand what work actually gets done at the task level. Identify which tasks AI can handle well today, which ones it might handle in a year, and which ones require human judgment for the foreseeable future.
Invest in your people. Give them tools, training, and time to learn how to work alongside AI. The organizations that will win the AI era aren’t the ones with the fewest employees. They’re the ones with the most capable employees.
Measure output, not headcount. If AI makes your team 30% more productive, you don’t need 30% fewer people. You need 30% more ambition. Ship more features. Enter new markets. Solve harder problems. Grow into the capacity that AI creates.
The companies laying off thousands and calling it AI transformation will look back on this moment with regret. Not because AI didn’t work, but because they sacrificed the institutional knowledge they needed to make it work well.