Hiring data on AI-exposed young workers and startup founder surveys show a slower job market, not mass layoffs. Here's what's really happening.
Is AI actually taking jobs right now? #
The clearest available data does not show mass firings caused by AI. It shows a hiring slowdown concentrated among young workers in occupations considered highly exposed to AI. One measure found that employment among workers age 22 to 25 in highly AI-exposed jobs ran about 19% below where it would have landed if it had kept pace with less-exposed occupations. That gap came from companies hiring fewer young workers into those roles, not from layoffs. It’s a real effect, but it’s a narrower and more ambiguous story than “AI is replacing workers.”
TL;DR #
Hiring, not firing, is where the AI effect shows up, with young workers in AI-exposed occupations seeing employment run roughly 19% below trend, driven by reduced hiring rather than job cuts.The divergence predates ChatGPT, which makes it hard to say how much of the current gap is AI-caused versus a continuation of a trend that started before generative AI existed.Founders are using AI heavily but rarely say it’s essential, with a Gusto survey of over a thousand 2025 startup founders finding 60% used AI at launch, half said it sped things up or cut costs, and only 3% said they likely wouldn’t have started the company without it.Businesses treat AI as augmentation, not replacement, because most jobs require accountability and judgment that a model can’t yet carry with real consequences on the line.** Intelligence being cheap doesn’t automatically mean fewer employees**, because running AI at scale often multiplies the number of model calls per task rather than simply substituting for a person.The bigger economic story right now is financing AI infrastructure, not labor displacement, and that financing boom is what determines whether AI capacity keeps expanding at all.
Remy is new. The platform isn't. #
Remy is the latest expression of years of platform work. Not a hastily wrapped LLM.
What does the hiring data actually say? #
The strongest data point on AI and employment right now comes from looking at young workers, ages 22 to 25, in occupations classified as highly exposed to AI tools. Employment in that group came in about 19% below the level it would have reached if it had simply tracked employment trends in less-exposed occupations. That’s a meaningful gap.
But the mechanism matters. This wasn’t a wave of layoffs. It was a slowdown in hiring. Companies weren’t pushing existing young employees out the door. They were bringing fewer new ones in. That distinction matters for anyone trying to figure out whether AI is a threat to jobs that already exist versus a bottleneck for people trying to get their first ones.
It also matters that this divergence between AI-exposed and less-exposed occupations started before generative AI tools existed in their current form. That timing doesn’t rule out an AI effect. It does mean researchers can’t cleanly attribute the entire gap to AI. Some of it could be a pre-existing structural trend that AI is now accelerating, or it could be something else entirely showing up in the same occupations that happen to be AI-exposed.
Why would companies hire less without firing anyone? #
The likely explanation is that entry-level and junior roles are often where AI tools substitute most directly for tasks a new hire would otherwise be assigned: drafting, summarizing, basic coding, first-pass research. If a senior employee can get through that work faster with an AI tool, the incremental need to hire a junior person to do it shrinks. That doesn’t require firing anyone. It just means the next open headcount doesn’t get filled, or gets filled more slowly.
This is a quieter and slower-moving effect than mass layoffs, and it’s harder to see in real time. It shows up in aggregate hiring statistics over months or years, not in a single company’s earnings call. That makes it easy to miss and easy to argue about, because by the time the pattern is clear in the data, it’s already been happening for a while.
Are founders actually replacing workers with AI? #
The startup data suggests something more modest than replacement. A Gusto survey of 1,051 people who started companies in 2025 found that 60% used AI tools during the launch process, and about half said AI made starting the business significantly faster or cheaper. Those are large numbers, and they show AI is now a standard part of how people build companies.
But only 3% of those founders said they likely wouldn’t have started the company at all without AI. That’s a telling gap. AI is widely used as an accelerant and a cost-cutter, but very few founders see it as the thing that made a business possible in the first place. It’s closer to a productivity tool than a replacement for the underlying work of building a company.
Why isn’t cheaper intelligence translating into fewer jobs? #
Other agents start typing. Remy starts asking. #
Scoping, trade-offs, edge cases — the real work. Before a line of code.
There’s a reasonable assumption that if AI models get good enough and cheap enough, companies will simply need fewer people to do the same amount of work. That assumption runs into two problems.
First, falling costs per unit of AI usage don’t necessarily reduce total spending or total AI usage. When token prices drop, organizations tend to run models more often, give them more steps to check their own work, or apply them to tasks that weren’t worth the cost before. A single completed task on the more advanced end can already involve dozens or hundreds of individual model calls. Cheaper tokens can expand how much AI gets used rather than simply shrinking the bill.
Second, and more directly relevant to hiring: most jobs aren’t just a bundle of tasks that can be swapped for a model’s raw output. They require someone to be accountable when something goes wrong, to exercise judgment in ambiguous situations, and to be present and responsible in ways that don’t reduce cleanly to “can an AI produce this text or this code.” Businesses evaluating whether to replace a role with AI tend to ask whether they can trust the outcome with real money and real consequences attached, not just whether the underlying task is technically automatable.
That gap between technical capability and organizational trust is a big part of why AI adoption has grown fast without producing a matching wave of job elimination. Companies are using AI as a tool employees are expected to use, layered on top of existing roles, rather than as a wholesale substitute for the people in those roles.
Is job security actually at risk from AI? #
The honest answer is: not evenly, and not in the way “AI is coming for your job” headlines suggest. The clearest risk shows up at the entry level, in hiring rather than firing, and in occupations already flagged as highly exposed to AI tools. That’s a real and worth watching, particularly for young workers trying to break into fields like software development, customer support, or basic analysis where AI tools now handle a meaningful share of junior-level tasks.
For workers already established in roles that require judgment, accountability, and coordination with other people, the data doesn’t show anything resembling a mass displacement event. Adoption of AI tools is high. Founders and firms are using them constantly. But the leap from “AI can do this task” to “we no longer need a person accountable for this work” hasn’t happened at scale, and the available data doesn’t support the idea that it’s happening quietly either. The bigger open question is how the AI industry’s own infrastructure buildout, financed increasingly through complex arrangements between chipmakers, cloud providers, and large pools of institutional capital, continues to lower the cost and raise the capability of these tools over time. If that buildout keeps going, the pressure on entry-level hiring in exposed occupations could grow. If it stalls, so does the pressure.
Frequently Asked Questions #
Is AI actually causing layoffs right now?
The available data doesn’t show a clear pattern of AI-driven layoffs. The strongest signal is reduced hiring for young workers in AI-exposed occupations, not job cuts among existing employees.
Everyone else built a construction worker.
We built the contractor.
One file at a time.
UI, API, database, deploy.
Which jobs are most at risk from AI?
Entry-level and junior roles in fields already flagged as highly AI-exposed show the clearest effect, mainly through slower hiring rather than displacement of current workers.
Do startup founders think AI is essential to running a business?
Not really. Most founders who used AI said it made launching faster or cheaper, but very few said they wouldn’t have started their company without it.
Why hasn’t cheap, capable AI led to mass job cuts?
Jobs require accountability and judgment that businesses are reluctant to hand fully to a model, and cheaper AI usage often expands how much AI gets run rather than simply cutting headcount.
Could this change as AI infrastructure keeps expanding?
It’s possible. The pace of AI capability and cost improvements depends heavily on continued infrastructure investment, and that could shift hiring patterns further, especially at the entry level.