There’s a clause appearing in job adverts that should make every professional pause — but the panic around it misses what’s actually happening.
I spotted it on a temporary contract advert recently: a final line, almost casual, noting that the successful candidate would be “helping to train an AI agent” as part of their day-to-day work. Read it twice and the implication lands — you’re being paid to teach the thing that may eventually do your job.
It’s an unsettling clause. But before anyone updates their CV in a cold sweat, let’s do what good career strategy always requires: separate the signal from the noise.
The paradox is real — and it’s a genuine business strategy
There’s now an enormous, fast-growing market for exactly this. Mercor, a company that recruits people to train AI models, has a CEO who openly predicts that “training agents is going to become the largest job category in the world.” The pay is real money, too — these roles average around $105 an hour, with deep domain experts (a psychiatrist designing clinical scenarios, say) commanding up to $350.
The logic for employers is simple and slightly ruthless. The most valuable thing an experienced professional owns isn’t their task output — it’s their tacit knowledge: the judgement, the pattern recognition, the “here’s how I actually decide” that lives in your head and never makes it into a process document. A short, well-scoped contract lets a company pay for that expertise and capture it into a model simultaneously. You leave; the encoded knowledge stays.
One screenwriter doing this work captured the moral knot perfectly. Some peers, she said, view it as “crossing the picket line.” Her counter: “Do you want AI to be good because it’s being trained by good people, or not?” That’s the paradox in a single sentence.
Now — the part the headlines skip
Here’s where cutting through the hype matters. Despite the alarm, the actual displacement looks very different from “the robots are coming for your desk.”
Of roughly 1.2 million US layoffs announced in 2025, only about 4.5% explicitly cited AI. And the displacement that is happening is concentrated in a specific place: entry-level hiring. As one 2026 analysis put it, AI “isn’t taking the job you already have; it’s taking the job you were going to be hired into.” It shows up as a quiet hiring slowdown at the bottom of the ladder, not a layoff wave for established professionals.
The cautionary tale here is Klarna. Its famous claim that an AI assistant did the work of 700 agents turned out to be a workload-equivalence calculation, not 700 redundancies — the company had already frozen hiring. By 2025 the CEO admitted they’d over-indexed on cost, the quality dropped, and Klarna was rehiring humans and repositioning live human support as a premium feature.
Why do incumbents hold up better? A Harvard/BCG study of 758 consultants found AI boosted quality by around 40% on tasks inside its competence — but made people 19 percentage points less likely to get the right answer on tasks just outside it, because the model was confidently, plausibly wrong. Almost every real job is a mix of both. That’s precisely why AI automates tasks far faster than it automates jobs.
The shift that actually matters for your career
PwC’s 2026 Global AI Jobs Barometer analysed over a billion job ads and found a two-track labour market emerging. Roles being “professionalised” by AI — reshaped to demand more human judgement — are growing twice as fast as roles being “democratised,” with 42% higher wage growth. Most tellingly: the most AI-exposed junior roles are now seven times more likely to demand traditionally senior skills like leadership.
The ladder isn’t disappearing. It’s compressing. The bottom rungs are thinning, and the skills that used to define seniority — judgement, stakeholder management, strategic framing — are being demanded earlier and rewarded more.
So what do you actually do? Six safeguards
- Read the contract clause as information, not a threat. If an advert says you’ll train an AI, that tells you exactly which tasks the company considers automatable. Treat it as free market intelligence about where not to anchor your value.
- Move up the judgement stack. The durable work is the “jagged frontier” stuff AI gets confidently wrong — ambiguous decisions, novel situations, reading the room. Deliberately concentrate your visible contribution there.
- Become the person who directs the AI, not the person it learns from. The new premium role is human oversight: framing problems, interpreting outputs, catching the plausible-but-wrong. AI literacy is now a baseline professional skill, not a niche one.
- Make your tacit knowledge a relationship, not a deliverable. The judgement that can’t be cleanly captured in a training session — built on trust, context and ongoing dialogue — is exactly the judgement that’s hard to encode and walk away from.
- Know your rights and watch for transparency. From January 2026, Ontario requires employers to disclose AI use in job adverts, and the EU AI Act treats many employment AI systems as high-risk. The direction of travel internationally is clear: AI-assisted employment decisions remain employment decisions, and employers stay accountable for them.
- If you take a training role, take it with eyes open. Some professionals reframe this work as adaptation — getting an insider’s view of how the technology learns, which is itself a hedge. That’s legitimate. Just price the contract knowing what you’re contributing, and don’t sign away your wider competitive edge for a single day-rate.
The honest bottom line
You are probably not being quietly engineered out of your own role. But the next version of your role — and the entry point juniors used to climb through — is genuinely being reshaped. The professionals who thrive won’t be the ones who refuse to touch AI, nor the ones who train it uncritically. They’ll be the ones who understand exactly what it can’t do, and make that the centre of their value.
