When journalists spin possibility into inevitability, they encourage panic-driven career decisions.
If you’ve seen headlines this month about “jobs AI will replace,” I have news: you’re being sold a story that even Microsoft (who conducted the study) explicitly warns against.
Over the summer, Microsoft released a report titled Working with AI: Measuring the Applicability of Generative AI to Occupations. This 42-page report includes two lists of jobs:
- Occupations with the most overlap with common Bing and Microsoft Copilot tasks (including translators, writers, historians, archivists, data scientists and web developers).
- Jobs with the least AI task overlap (phlebotomists, embalmers, dishwashers, roofers).
The report is detailed on its methodology and limitations, but nuance doesn’t drive clicks, so the headlines have turned it into:
“The 40 jobs most at risk from AI, and 40 it can’t touch” – Sky News[1]
“Microsoft reveals the 40 jobs AI is most likely to replace” – Tom’s Guide[2]
“Jobs most at risk from the AI boom revealed” – The Independent[3]
You’ve probably seen one of these articles. They’ve been everywhere for weeks now.
There’s only one problem. They’re all wrong.
What Does the Report Say?
The report itself states:
It is tempting to conclude that occupations that have high overlap with activities AI performs will be automated and thus experience job or wage loss, and that occupations with activities AI assists with will be augmented and raise wages. This would be a mistake…[4]
In other words, their data does not support the headline-grabbing simplification of “at risk” and “safe” jobs. The actual impact of AI use in any of the professions listed is subject to innumerable factors. In fact, some of these jobs may even experience a boom as a result of AI:
Prior research has shown that automation can either reduce or increase employment in an occupation, and the same can occur following the introduction of AI. Understanding these effects requires data and analyses that are beyond the scope of this work.[5]
For example, it was assumed that ATMs would replace bank tellers, but they didn’t. As it happens, they had the opposite effect – they made banks cheaper to operate, so banks opened more branches. More branches meant more tellers needed, and those tellers shifted from routine transactions to relationship building and complex problem-solving. The technology changed and even expanded the role; it didn’t eliminate it.
How This Distortion Happened
The ATM example comes from the Microsoft report itself. The researchers explicitly warn against the conclusions the media drew. But that caution is lost in the quest for attention-grabbing headlines.
This isn’t new. Ben Goldacre wrote about this 20+ years ago:
Once journalists get their teeth into what they think is a scare story, trivial increases in risk are presented, often out of context, but always using one single way of expressing risk, the “relative risk increase”, that makes the danger appear disproportionately large[6]
In this case, the paper discusses AI task overlap, but does not talk about jobs being replaced. The former is about technological capability, the latter is about economic outcomes. By combining these, journalists can create scary sounding headlines which use accurate data to draw a false conclusion – in this case, saying that particular jobs will or will not be viable in the near future.
It’s also worth noting that even if we take the report at face value, we are looking at a report from the company that makes Bing and Copilot, which has used data they’ve gathered from Bing and Copilot to conclude that Bing and Copilot are going to have a transformational impact on the entire world of work. This is hardly a shocking message.
The articles use Microsoft’s brand name as an appeal to authority – we trust it as a reputable name in the technology industry. But we must remember that when Microsoft says “AI can do 70% of a translator’s tasks”, the implicit message is “buy our translation AI”.
So we have: researchers who explicitly warn against drawing job displacement conclusions, a corporation with an incentive to promote AI adoption, and media outlets that turned both into “these careers are doomed”. Let’s examine what that looks like for the #1 job on their list.
Case Study: Translation
To illustrate why the study’s caution matters, consider two equally plausible futures.
In our first potential future, let’s say the headlines are right. 98% of a translator’s tasks can be automated, so there are mass job losses to AI. Learning languages becomes redundant as speech and text can be accurately translated in seconds by machines, so all translators and interpreters need to retrain and find new jobs.
In our second potential future, let’s bring in a couple more perspectives. If you don’t work in translation, it’s tempting to think that it’s just an act of swapping out one language for another. But as with all creative and professional endeavours, it is of course more complex than that:
Translation is not an act of mechanically swapping words between languages. It’s an act of cultural mediation, especially with fiction. A translator has to know not just what the author wrote but what the author meant, and how that meaning will be received by an audience with a very different set of assumptions, idioms, and taboos.[7]
We can see this in action if we look at multiple translations of the same text. For example, Emily Wilson’s translation of The Odyssey (2017) has the opening line, “Tell me about a complicated man” rather than the usual translation of “a man of twists and turns”. This fundamentally changes the entire story – now Odysseus isn’t just a passive victim of fate, but someone whose inner conflict is central to the plot. Same origin text, substantively similar translation, but a totally different story.
Given this complexity, here’s an alternative: AI supercharges translation rather than decimating it. By making translation quicker and easier, it opens a whole world of media (films, TV, novels, plays, self-help books, industry publications, investigative journalism, the list is endless). Demand for good, context-informed translation goes through the roof, and good translators have access to tools that enable them to work on more texts, at greater pace, than ever before. Translation becomes a boom industry.
The Microsoft report explicitly warns against assuming either future is more likely than the other, but the media version would have you believe the worst-case scenario is inevitable.
Career Implications
We’ve established that the real story is starkly different than the headlines. So what? Why does this matter for career planning?
It matters because reporting like this can cause us to make decisions without any basis in reality. It’s absolutely logical to look at a list of jobs that are going to be redundant in the future and deliberately avoid them, or to look at the list of “future-proof” ones and gravitate towards those. But if we use labour market information which is built on poor foundations, we’re not making informed decisions – we’re being led astray.
How many budding translators, historians, writers, data scientists, geographers or archivists are deterred by these headlines? How many people with little aptitude or interest are now considering careers in phlebotomy or nursing or hazardous waste removal because they’ve been told these jobs will be secure into the future?
The headlines seem instructive, but the decisions they’re encouraging us to make are not supported by any of the data they’re reporting on.
Now that we know all that we know, let’s take a more reasoned response:
Don’t conclude:
- Career X is doomed, avoid it
- Career Y is safe, pursue it
- We know which jobs will exist in 10 years
Do conclude:
- Some roles will see more obvious AI integration than others
- The relationship between technology and work is complex and set to become more so
- Adaptability matters more than picking the “right” field
- Career decisions should be based on your interests and abilities, not panic-driven predictions
And if you’re genuinely concerned about AI’s impact on your field, the best response isn’t avoidance, it’s engagement. Understand what AI tools can actually do in your domain. Experiment with them and think critically about their utility. Consider whether you could develop skills in directing, evaluating, and contextualising AI output.
Whatever happens with AI, one thing is constant: the professionals who thrive won’t be those who picked the “safe” job from a Forbes listicle, they’ll be those who followed their values and passions, and learned to work effectively in a changing landscape.
Conclusion: Truth vs Sensationalism
Next time you see a headline about “jobs AI will replace,” ask yourself: does this come from the study authors, or from the headline writers? The media translated “high task overlap” into “career extinction”. In doing so, they lost something crucial: the nuance, the complexity, and ultimately, the truth.
[1] https://news.sky.com/story/the-40-jobs-most-at-risk-of-ai-and-40-it-cant-touch-13447013
[2] https://www.tomsguide.com/ai/microsoft-reveals-the-40-jobs-ai-is-most-likely-to-replace-and-40-that-are-safe-for-now
[3] https://www.independent.co.uk/news/uk/home-news/ai-jobs-risk-microsoft-uk-layoffs-redundancies-b2844802.html
[4] https://arxiv.org/pdf/2507.07935
[5] ibid
[6] https://www.theguardian.com/science/2005/sep/08/badscience.research
