Every few weeks the question of whether AI is about to gut white collar work flares up again. This time the spark is an analogy. A viral essay from AI builder Matt Shumer compared the current moment to February 2020, the eerily normal weeks just before the pandemic rewrote the economy, arguing that the warning signs about AI and employment are being waved away in much the same manner. The comparison struck a nerve because the people running the largest AI labs keep saying alarming things, then softening them.
The predictions keep escalating
The headline forecasts are not subtle. Microsoft AI chief Mustafa Suleyman told the Financial Times that most tasks done by lawyers, accountants, project managers and marketers could be fully automated within 12 to 18 months. Anthropic's Dario Amodei has warned that AI could eliminate more than half of entry level white collar jobs within five years. Ford chief executive Jim Farley said AI would replace literally half of all white collar workers in the United States. OpenAI's Sam Altman has gone further still, conceding that not even a chief executive's job, including his own, is ultimately safe.
Why the analogy resonates
What makes the February 2020 framing land is the gap between rhetoric and visible reality. Most offices still look much as they did a year ago. Hiring has slowed but not collapsed. To many workers the warnings feel abstract, in the way a distant outbreak felt abstract before lockdowns arrived. Shumer's point is not that catastrophe is certain but that people are poor at pricing in non linear change, and that the same leaders sounding the alarm have a commercial interest in both hyping and downplaying it depending on the audience.
The data is starting to move
The debate is no longer purely speculative. Challenger, Gray and Christmas now lists AI as the leading reason companies give for cutting jobs, with AI cited in around 40 percent of cuts announced in May, up from 7 percent in January. AI linked reductions reached 87,714 for the year through May, more than the previous two years combined. A Bloomberg opinion piece this week argued that the real AI jobs crisis is the one not being discussed openly, the slow disappearance of the entry level roles that once let people break in. The numbers are still modest against a workforce of millions, but the direction is consistent.
The case for scepticism
There is a strong counter argument, and it is not just denial. Critics point to AI washing, the habit of attributing ordinary cost cutting to a futuristic cause that sounds better to investors than admitting to overhiring. Labour economists note that previous waves of automation anxiety, from spreadsheets to offshoring, produced disruption without the mass unemployment the loudest voices predicted. The honest position is uncomfortable: the warnings could be early rather than wrong, or they could be marketing dressed as prophecy. February 2020 is a potent image precisely because nobody knew, at the time, which it would turn out to be.
What to actually do about it
For workers, the practical takeaway is less about predicting the date than about hedging. The roles proving most durable are those that combine domain judgement with the ability to direct AI tools rather than compete with them. For employers, the temptation to treat AI as a quiet headcount lever carries its own risk, hollowing out the junior pipeline that senior roles depend on. The value of an essay like Shumer's is not its accuracy but its prompt: if this really were February 2020, what would you want to have done before March arrived?
