When a business leader asks what AI will do to work, the usual answers are about tasks, tools and headcount. A Guardian investigation published on 7 April 2026 offers a different and more uncomfortable answer: for a growing group of experienced professionals, AI has already become the job they take when nothing else is left. The piece, reported by The Guardian, profiles five skilled workers aged 50 and older who now train the very models that may one day do parts of their old jobs. For anyone deciding how to deploy AI at work, it is worth reading closely — not as a morality tale, but as a labour-market signal.
What the reporting actually shows
The work in question is data annotation: labelling and evaluating the information used to train models such as OpenAI’s ChatGPT or Google’s Gemini. A doctor might review how a model answers medical questions, flag unsafe or incorrect responses, and suggest better ones. The Guardian names companies operating large contractor networks for this work — Mercor, GlobalLogic, TEKsystems, micro1 and Alignerr — with clients including OpenAI, Google and Meta, as well as academic researchers and industries such as healthcare and finance.
Refer & earn: 30% of a client’s first month, then 10% of every month after that, for up to 24 months per client. See the partner programme →
Pay is wide-ranging. According to online job postings cited in the article, AI training gigs start at $20 an hour, rising to between $30 and $40 an hour, and in some cases trainers with coveted subject-matter expertise can earn over $100 an hour. The article also notes that top experts can, in some cases, earn over $180 an hour. But it is contract-based: unstable hours, no benefits, and a clear step down from the six-figure, benefits-bearing roles that professionals in software development, medicine or finance typically held, according to the US Bureau of Labor Statistics.
The human stories make the economics concrete. Patrick Ciriello, 60, holds a master’s degree in information management and spent most of his career designing software systems for banks, universities and pharmaceutical companies. After losing a job at the start of 2023, he sent out hundreds of applications — IT support, customer service, even a deli counter role — and got no offers. His family spent about four months sleeping in a Toyota Highlander. A cryptic LinkedIn message in March 2024 led to a content-writing job that turned out to be AI training. His first such role paid $21 an hour; after a mass layoff in January 2025, he found another AI training role, evaluating model responses for Meta since August 2025, at $20 an hour. His earnings are low enough that he qualifies for Medicaid and SNAP benefits.
Rebecca Kimble, 52, spent more than a decade as an emergency medicine physician, earning between $300,000 and $500,000 a year. After a DUI placed her on administrative leave, then a breast cancer diagnosis and more than two years away from clinical work, she found re-entry blocked. The American Medical Association notes that extended gaps in practice make it difficult for physicians to return, as hospitals prioritise actively practising candidates. By January, she was juggling AI training assignments on several platforms, evaluating how models responded to medical questions.
The structural backdrop
The article does not present these as isolated misfortunes. It cites Richard Johnson, vice-president of the AARP Public Policy Institute, who researches how age bias affects job outcomes: US workers over 60 take about 50% longer to find new jobs than people in their 20s and 30s, and only a fraction regain their previous earning levels. Employers may erroneously view older workers as more expensive, less current in their skills, and harder to train.
It also cites the Urban Institute, a policy research non-profit, finding that about half of US workers aged 50 to 54 are involuntarily pushed out of long-term jobs before they expect to retire, and the Economic Policy Institute, which reports that roughly 5.7 million workers over 55 lost their jobs in early 2020, with many yet to return to stable work.
Joanna Lahey, a professor at Texas A&M University who studies age discrimination and labour outcomes, describes “bridge jobs” — lower-paying, less demanding roles that help workers stay afloat as they approach retirement. Historically that meant temp work, retail, fast food, or gig driving. Now, for engineers, lawyers, nurses and designers, AI data training is becoming the new bridge job. Lahey told the Guardian:
[AI] training work may be better in some ways than those earlier alternatives.
That is a fair point. The work can offer flexibility, quick income and intellectual engagement. It is also, as the article states plainly, often a step down — and in some cases it means training the systems that may eventually replace people in the very professions the trainers came from.
Why this matters to a business reader
Three practical implications stand out for anyone deciding what to do about AI at work.
First, the AI supply chain is human, and it is not always treated as a workforce. The models your business buys or builds are shaped by contractors reviewing outputs, flagging errors and writing better answers. If your organisation is procuring AI, it is also, indirectly, procuring labour. That raises questions worth asking vendors: who annotates the data, under what contract terms, and with what quality controls? The Guardian’s reporting shows this layer is large, distributed across many firms, and staffed substantially by experienced professionals working without benefits. That is a resilience question as much as an ethical one — contract-based annotation capacity can be cut quickly, as Ciriello’s January 2025 layoff illustrates.
Second, expertise is the scarce input, and your organisation may already be sitting on it. The reason a physician or a software architect is valuable to an AI trainer is the same reason they are valuable to you: they can tell a wrong answer from a right one. Businesses planning AI rollouts often focus on tooling and ignore the internal review capacity that makes outputs trustworthy. The Guardian’s subjects are effectively doing that review work externally, for platforms, after being unable to find roles in their fields. Employers that treat experienced staff as a liability rather than as evaluation capacity may be discarding the exact judgement their AI systems need.
Third, the bridge-job framing should inform workforce planning. If AI training is becoming the default landing spot for displaced experienced professionals, then the transition costs of AI adoption are being absorbed by individuals and public programmes — Medicaid and SNAP in Ciriello’s case — rather than by the organisations capturing the productivity gains. That is a reputational and, increasingly, a regulatory exposure. It is also a retention risk: the same demographic is a deep source of institutional knowledge.
What to do about it
- Audit your AI supply chain. Ask vendors who does the annotation and evaluation, how those workers are engaged, and what happens to quality if that capacity is cut.
- Build internal evaluation capacity deliberately. Subject-matter experts are the quality control layer for AI outputs. Treat their review time as a funded responsibility, not a favour.
- Review age assumptions in hiring and redeployment. The bias Johnson describes — older workers seen as costlier and less trainable — is a documented pattern, and it sits awkwardly beside the fact that AI trainers are being hired precisely for the expertise that comes with experience.
- Plan for transition, not just deployment. If roles will shrink, say so early and invest in redeployment into the review, oversight and exception-handling work that AI systems create.
None of this is an argument against AI adoption. It is an argument for seeing the whole picture. The Guardian’s reporting shows that the labour market around AI is already mature enough to have a bottom rung — and that the people on it are often the ones with the most to teach the models. Businesses that notice this early can make better procurement, workforce and risk decisions than those that treat AI as a purely technical project.
Faida Workforce advises organisations on workforce planning and AI adoption, including the human review layers that make AI outputs reliable. We mention this because it is relevant to the decisions above; readers should weigh it as they would any supplier’s perspective.
Sources
- ‘There’s a lot of desperation’: skilled older workers turn to AI training to stay afloat — The Guardian, published 7 April 2026.
By Faida Workforce Editorial. AI tools assist in drafting and research; a human editor reviews and is accountable for the published piece.
30% of their first month, then 10% every month after that
Refer a business to Faida Workforce and you earn 30% of the client’s first paid month, then 10% of every month for up to 24 months per client — rising to 15% from the accrual after you pass 10 active clients. Paid by PayPal, batched on the 1st.
10 Pro clients = $2,598.70 over 24 months — $3,748.13 at the 15% tier.
Figures are potential earnings worked from the published rates, assuming every referred client pays for the whole cap — not a promise, not an average. The $1 trial is never commissionable.
