I want to defend a position that sounds like a cop-out but isn’t: most AI product launches will change what’s expected of you long before they change whether you have a job.
This isn’t optimism. It’s not “AI won’t take jobs.” It’s a claim about sequence. The first-order effect of an AI product launch is not a layoff. It’s a quiet rewrite of the job description — the tasks you’re supposed to do, the speed you’re supposed to do them, and the skills you’re supposed to have.
The layoffs come later, if they come at all. And by the time they do, the job has already changed so much that “eliminated” and “redefined out of existence” are hard to tell apart.
I’ve spent months collecting the evidence on this. Here’s my case. Tear it apart.
First, Let’s Be Precise About the Claim
“Change job expectations before they eliminate jobs” can mean several things. Let me define what I mean — and what I don’t.
What I’m claiming:
When a company deploys an AI product, the first measurable effect is a change in task composition and performance expectations, not headcount.
This happens because automating a task is easier than automating a job. Most jobs are bundles of tasks. AI eats the routine ones first, leaving the judgment-heavy ones. The job survives, but it becomes a different job.
Layoffs, when they happen, usually follow the expectation change — either because the company decides it needs fewer people to do the new version of the job, or because workers who can’t make the transition leave and aren’t replaced.
What I’m not claiming:
That AI won’t eliminate jobs. It will, in some cases.
That the expectation change is harmless. It often isn’t.
That companies are being honest when they frame AI adoption as “augmentation.” Some are, some aren’t.
The distinction matters because the public conversation conflates “AI will take your job” with “AI will change your job.” The first is dramatic and measurable. The second is quiet and harder to see — but it’s happening first, and it’s happening everywhere.
The Evidence: Expectation Change Is Already Here
The strongest evidence for my claim comes from the gap between what companies say about AI and what workers experience.
BCG’s fourth annual AI at Work report, based on a survey of 11,749 workers across 14 markets, found that 72% of respondents say AI has already considerably changed skills expectations in their roles. Nearly half — 47% — report spending more time managing and directing AI than doing the work itself. And 65% of managers and leaders believe agents will take over at least half of their job in the next three years.
Read that last number again. Two-thirds of managers think AI will handle half their job. They’re not predicting their own layoff. They’re predicting a fundamental shift in what their job is.
Korn Ferry’s Workforce 2026 Global Insights Report, based on a survey of more than 16,000 employees, found that 52% of workers say AI tools have increased the number of tasks expected in their role. 62% say their workload has increased significantly in the past two years. And 63% say AI has increased their efficiency — which, in practice, means they’re expected to do more, not less.
This is the pattern I’m describing. The tool arrives. The expectations change. The hours don’t.
Gartner’s survey of CHROs found that 22% report at least one business leader has stopped hiring for entry-level roles due to AI automation. But Gartner’s framing is important: the issue isn’t that entry-level jobs are disappearing. It’s that the tasks traditionally assigned to entry-level workers — routine execution — are being automated, leaving a mismatch between traditional skill profiles and the organization’s remaining lowest-complexity work.
The entry-level job isn’t gone. It’s been redefined. And the people who used to fill it don’t yet have the skills the new version requires.
The Hidden Mechanism: Task Recomposition, Not Job Elimination
Here’s the mechanism I think is doing most of the work — and it’s the part that gets least attention.
Most jobs are not single tasks. They’re bundles of tasks with different levels of routine and judgment. A customer support agent handles password resets (routine), account updates (routine), escalations (judgment), and emotional de-escalation (judgment). A marketing coordinator drafts copy (routine), schedules posts (routine), analyzes campaign performance (judgment), and decides what message to lead with (judgment).
AI product launches target the routine tasks first. They automate the password reset, the first draft, the data pull, the template fill. The judgment tasks remain. The job survives. But the ratio changes.
Research on automation pathways supports this. A 2026 study in Humanities and Social Sciences Communications found that AI-driven automation sustains occupational roles through emerging complementarity rather than substitution — meaning AI and humans end up doing different parts of the same job, not the same parts at different speeds.
The Denmark study — the most rigorous labor market analysis of AI chatbots to date — found precisely estimated null effects on earnings and recorded hours across 25,000 workers in 11 exposed occupations, two years after ChatGPT’s launch. But the same study found that adoption is linked to occupational switching and task restructuring.
That’s the pattern in one sentence. No wage effect. No hours effect. But significant task restructuring. The job changes shape without changing its formal existence.
Why Expectation Change Precedes Layoffs
If the evidence shows task restructuring without headcount reduction, why do I say the expectation change comes before the layoff — rather than instead of it?
Because the layoff, when it comes, is usually a second-order decision made after the expectation change has already happened.
Here’s the sequence I’ve seen:
Phase 1: Tool deployment. The AI product launches. Workers are told it’s a tool, not a replacement. The tool handles routine tasks. Workers adapt.
Phase 2: Expectation reset. As the tool proves capable, performance expectations shift. “You should be able to handle 30% more tickets now.” “Why isn’t the report done faster?” The job description hasn’t changed, but the bar has.
Phase 3: Capacity assessment. Leadership looks at the new productivity numbers and asks: “If we can do this with the current team, do we need the same team?” This is the moment when layoffs become thinkable. Not because the AI replaced the job, but because the expectation change made the job look different enough that the old headcount looks excessive.
Phase 4: Either redefinition or reduction. Some companies redefine roles — retraining workers for the judgment-heavy tasks AI can’t do. Others reduce headcount, assuming the remaining workers can absorb the rest. The data suggests both happen, and the second is more common.
Gartner’s research confirms the second phase is real — and the fourth is often a mistake. The firm predicts that by 2029, 30% of employees laid off due to AI replacement will need to be rehired, often at significantly higher cost. “Business and IT executives who use AI primarily as a tool for cost cutting risk making reductions that are too deep and too soon,” Gartner warns.
The rehire prediction is the strongest evidence for my claim. If AI were genuinely replacing jobs — automating the work, not just changing the expectations — companies wouldn’t need to rehire. They’d just keep the automation. The fact that Gartner expects a significant rehire wave means companies are discovering that the expectation change broke something the automation couldn’t fix: institutional knowledge, judgment, and the human relationships that hold workflows together.
The Layoff Headlines Are Misleading
I want to address the counterargument directly, because it’s the strongest one: “But look at the layoffs. 180,000 jobs linked to AI since 2023. 112,000 in 2026 alone.”
Those numbers are real. Challenger, Gray & Christmas has tracked more than 180,000 corporate job losses attributed to AI since May 2023. In July 2026, AI was the most common reason cited by companies for job cuts for the fifth month in a row.
But here’s what the FT’s analysis found: companies have strong incentives to attribute layoffs to AI rather than to more mundane reasons like poor financial performance, restructuring, or shifting investment priorities. Meta’s 1,000 job cuts in January 2026 came from its metaverse division — the company was redirecting funding toward AI devices, not replacing workers with AI.
Carl-Benedikt Frey, associate professor of AI and work at the Oxford Internet Institute, put it bluntly: “I haven’t seen any compelling evidence of the narrative that these firms are automating a lot of work. I don’t think AI-driven automation is the key story here”.
The FT also found that fewer than half the AI-related job cut announcements resulted in a share price boost — and companies citing AI as a factor underperformed the Nasdaq by almost 10% in the following 30 trading days. If AI layoffs were a sign of genuine efficiency, you’d expect the market to reward them. It doesn’t consistently do so.
This matters for my argument because it suggests the layoff narrative is overstated. The expectation change — the quiet, invisible part — is the bigger story. And it’s happening whether or not the layoffs materialize.
The Counterargument: Where I Might Be Wrong
I want to steelman the opposite view, because I don’t think my claim is airtight.
Counterargument 1: Some jobs really are being eliminated, not redefined. Klarna cut 700 support positions between 2022 and 2024, replaced them with AI, then had to rehire humans when quality dropped. Ford rehired 350 veteran engineers after an AI-driven design experiment failed. These are not expectation changes. They’re failed replacements. The job disappeared, then came back. If the rehires hadn’t happened, the job would be gone.
Counterargument 2: Entry-level roles are being cut, not just redefined. Gartner found 22% of CHROs report entry-level hiring stopped due to AI. Revelio Labs found entry-level postings down 35% in 18 months. If companies stop hiring for a role category entirely, that’s not expectation change. That’s elimination — gradual, attrition-based, but elimination nonetheless.
Counterargument 3: The expectation change is itself a form of elimination. If a job becomes 30% harder, 20% more judgment-heavy, and requires skills the current holder doesn’t have, does it matter whether the title survives? The person who can’t make the transition is effectively replaced, even if the headcount number doesn’t change.
Counterargument 4: The Denmark study may not generalize. Denmark has strong labor protections, high union density, and a generous social safety net. The null effects on wages and hours might reflect Danish labor institutions, not a universal pattern. In the U.S., where employment-at-will is the default, the same productivity gains might translate to layoffs more quickly.
These are real weaknesses. I don’t have a perfect answer to any of them. But I think they refine the claim rather than refute it. The expectation change is still the first-order effect. It’s just that in some cases, the expectation change is severe enough that the job effectively disappears — even if the title doesn’t.
What This Means for Workers
If my claim is right — that expectation change precedes elimination — then the most important thing a worker can do is track the expectation change, not the layoff headlines.
Here’s what I’m doing:
Mapping my tasks against the AI tools in my workflow. Not “is my job safe?” but “which tasks are being automated, which are being augmented, and which are genuinely mine?” The answer tells me where the expectation change is heading.
Tracking my workload, not just my hours. If my output is up 20% and my pay is flat, that’s the expectation change in action. I want data, not vibes.
Watching for the capacity assessment. When leadership starts talking about “doing more with the same team,” that’s Phase 3. The layoff conversation, if it comes, happens after that.
Building evidence of the judgment work. The tasks AI can’t do are the ones I want to be visibly good at. Not because they’re safe forever — but because they’re the ones that get me included in the redefined job, not excluded from it.
Asking the uncomfortable question. “If this tool changes what’s expected of me, how will we decide what my role looks like next?” Most managers don’t have an answer. But asking the question puts it on the record.
What I’m Still Wrestling With
I don’t have this figured out. Here’s what I’m stuck on:
1. How do you measure expectation change? Layoffs are countable. Expectation changes are qualitative. Is there a metric — quota increases, task reclassification, skill requirement shifts — that captures this before the layoffs happen?
2. Is the U.S. different? The Denmark study shows null effects in a high-protection labor market. Does the same pattern hold in the U.S., or does at-will employment accelerate the translation from expectation change to headcount reduction?
3. What happens to the people who can’t make the transition? Some workers are excellent at routine execution and struggle with judgment-heavy, ambiguous work. If the expectation change requires a skill they don’t have, the job is effectively gone for them — even if the title survives.
4. Is “redefined out of existence” a cop-out? I keep coming back to this. If the job changes so much that the original holder can’t do it, does it matter whether we call it “elimination” or “redefinition”? Maybe the distinction only matters to people who want to avoid saying “AI took jobs.”
5. What’s the strongest counter-evidence? I’m looking for it. If you have data showing that expectation change and layoffs happen simultaneously — or that layoffs come first — I want to see it.
No feedback yet — submit the first.