Every feed is full of it right now. AI is destroying jobs. AI will unlock a golden age of software creation. Entire industries are being hollowed out. Or we’re on the cusp of the most productive era in the history of technology. Pick your camp.
The strange thing is, both sides are citing the same headlines as evidence. The same announcements. The same numbers. And that tells you something important, not about AI, but about how badly we’re reading the moment.
We’re not watching one story unfold. We’re watching two. And we’ve decided, incorrectly, that they’re the same.
The template gets set
In early March 2026, Block announced it was cutting 4,000 jobs, roughly 40% of its workforce. CEO Jack Dorsey cited AI as the driving force. The company’s stock jumped 20%.
That investor reaction is the part worth sitting with. A brutal round of redundancies, and the market rewarded it.
What got validated that day wasn’t just Block’s strategy. It was a playbook. Cut headcount, cite AI efficiency, watch the stock move. Other CEOs noticed. The template was set.
The tell hiding in plain sight
Here’s where the AI-washing narrative gets its teeth.
When New York introduced a requirement for employers to cite a legal reason for mass layoffs, 160 companies filed notices. Not one of them, including Amazon and Goldman Sachs, both vocal about AI-driven efficiency elsewhere, checked the box marked “technological innovation or automation.”
Amazon’s own CEO Andy Jassy had previously pointed to generative AI and AI agents as workforce reduction drivers. He later walked it back: the cuts were “not really AI-driven, not right now at least.”
Klarna shed 40% of its workforce and framed it as an AI-motivated hiring freeze. Duolingo announced it would stop using human contractors for any work AI could handle. Bold moves, compellingly narrated.
But only 2% of executives surveyed said they made large staff reductions as a result of actual AI implementation. A further 60% admitted they did so in anticipation of AI efficiencies, speculating on a future that hasn’t arrived yet. Perhaps most revealing: 60% of hiring managers said they emphasise AI’s role in workforce decisions because it’s viewed more favourably than admitting financial pressure.
These aren’t AI stories. They’re overhiring hangovers, margin corrections, and capital efficiency drives, dressed up in the language of technological transformation. Tech companies tripled headcount between 2019 and 2023 on cheap capital and growth-at-all-costs mandates. When capital got expensive and margins started to matter, the reckoning was inevitable. AI arrived just in time to carry the narrative.
But the productivity shift is real
None of which means the efficiency gains are fiction.
Cash App’s design team went from 36 production pulls per month to over 1,000. That’s not a rounding error. Something structurally changed about what a team of that size can produce.
Salesforce eliminated 4,000 customer support roles and pointed to genuine efficiency gains from its agentic AI products. Duolingo’s AI-first hiring policy, whatever you think of how it was communicated, reflects a real change in what’s achievable without scaling headcount.
The productivity shift is happening. The question has never been whether AI changes what’s possible. It’s what companies choose to do with that capability when it arrives.
And that’s where the two futures diverge.
Two futures, one set of headlines
Here’s the actual fork in the road that the AI jobs debate keeps obscuring.
Some companies will use the leverage AI creates to shrink: cut headcount, reduce costs, improve margins, and call it transformation. The productivity gain becomes a financial manoeuvre, not a strategic one. The company does less with fewer people and presents it as progress.
Other companies will use the same leverage to build more than was ever possible before: more products, more scope, faster iteration, bigger ambition. Teams that previously couldn’t afford to explore a new product line now can. Engineers who spent 60% of their time on maintenance can spend 60% of their time creating. The productivity gain becomes a ceiling-remover, not a headcount calculator.
Both paths are being walked right now. Both cite AI. Both look, from the outside, like forward-thinking organisations embracing the future.
We are reading them as the same story. They are not.
How to read the signals
The press release won’t tell you which future a company is building towards. The product roadmap will.
The tell is in what happens to the space AI creates. Three outcomes are playing out right now:
- Headcount growing, ambition exploding. New products, new verticals, bets that were previously too expensive to take. AI didn’t replace anyone, it created the conditions to do more.
- Headcount flat, scope expanding. The same team shipping what used to require a team twice the size. The productivity gain gets reinvested into the roadmap.
- Headcount down, ambition unchanged. A smaller company maintaining the same product surface. The gain gets harvested as margin, not momentum.
The first two are where the interesting companies live. The third is the doom loop, dressed up as a transformation story.
Watch what companies do with the engineers who remain. Are they being given harder, more interesting problems? Or are they being asked to maintain the same roadmap with half the team? Watch the product announcements over the next 24 months. Are these companies shipping more, or are they shipping the same amount, just cheaper?
The 108,000 tech jobs cut in January 2026 alone, the highest monthly total since 2009, will look very different in hindsight depending on which companies those were. Some of those cuts will turn out to have been the precursor to something genuinely extraordinary. Others will just be the beginning of a long decline.
Right now, the headlines make them all look the same.
The choice being made right now
The doom loop and the golden age aren’t competing predictions about what AI will do to the industry.
They’re competing choices about what leaders will do with it. And those choices are being made right now, quietly, in roadmap reviews and hiring plans and conversations about where to reinvest the productivity gains that AI is already creating.
The companies that use AI to build more than they ever could will define what software looks like in the next decade. The companies that use it to justify doing less will spend that decade watching others do it.
The AI jobs debate asks the wrong question. It’s not whether AI will take jobs. Some of it will. Some of it already has. The question is what your organisation is building with the headroom it creates.
That answer won’t show up in a press release. But it will show up in the product.