Here’s a pretty provocative prediction: Microsoft AI CEO Mustafa Suleyman predicts “most, if not all” white-collar tasks will be automated by AI within 18 months.
Microsoft’s AI CEO is joining a chorus of executives who say they anticipate widespread job automation driven by artificial intelligence.
Mustafa Suleyman, the Microsoft AI chief, said in an interview with the Financial Times that he predicts most, if not every, task in white-collar fields will be automated by AI within the next year or year and a half.
“I think that we’re going to have a human-level performance on most, if not all, professional tasks,” Suleyman said in the interview that was published Wednesday. “So white-collar work, where you’re sitting down at a computer, either being a lawyer or an accountant or a project manager or a marketing person — most of those tasks will be fully automated by an AI within the next 12 to 18 months.”
The CEO said the trend is already observable in software engineering, in which employees are using “AI-assisted coding for the vast majority of their code production.”
“It’s a quite different relationship to the technology, and that’s happened in the last six months,” he said.
I really, really doubt that. AI has been able to do surprisingly well on a number of programming tasks mainly because it was created by programmers. It’s no wonder that it should be good at something like, say, creating a program to ingest JSON REST data into an SQL database. But the further you get from technical domains, the further you’re getting from people who can correct the AI when it gets things wrong.
Not every white collar employee is absolutely necessary, but as a class they possess the vital wells of institutional knowledge that lie beyond the datasets AI have been trained on. They also understand exception handling, knowing what to do when things go wrong. AI systems generally do poorly when faced with input combinations they’ve never seen before, which is why you have those videos of dozens of Waymo taxis blocking roads in clumps.
Also, given how zealously IT teams work to secure company data, are they really going to just blithely set AIs loose in their databases? Especially when their jobs may be among the ones target for elimination? Especially when AI is still prone to notorious hallucinations?
There are places where where AI might replace experts, namely those that use wide but highly structured datasets for narrow decision points, like some areas of regulatory law. But are decision makers really going to remove the ability to blame underlings for mistakes? “Sure we lost $100 million, but the AI told me it was OK!” is probably not going to wash as an adequate ass-covering maneuver. And, as I noted before, who is going to put an AI in charge of Accounts Payable when a single glitch could drain your entire bank account?
Plus there are entire classes of white collar jobs (sales comes to mind) where I don’t see AI making any headway replacing the fleshy incumbents.
Second, even where AI might effectively replace some humans, I don’t see companies letting Microsoft AIs in the front door. Copilot is a product people hate so much that current users won’t even turn it on even though they’re getting it free. Maybe Microsoft will make money getting in the back door via investments in OpenAI and Anthropic, but I doubt that’s going to reflect glory on Mustafa Suleyman.
But finally, let’s assume that his prediction is true and that most white collar jobs will be automated out of existence in the next 18 months. Just how does Microsoft expect to profit in the inevitable widespread economic collapse? If you think New York City’s trajectory is dire under Mandami now, how much direr is it going to get when vast swathes of people who pay the most taxes (and buy the most stocks) lose their income almost overnight? The subprime meltdown will look like a cakewalk in comparison when everyone starts selling their investment holdings just to pay for food and rent.
And if it’s known that the big AI companies are the direct causes of their immediate penury, how are such grandees going to protect themselves from the fury of the newly unemployed mobs? There were probably just a few thousand of Ned Ludd’s boys smashing mill machinery in early 19th century England. If Suleyman’s prediction came to pass, there would be what, 30 or 40 million people hurled into unemployment at the same time? In that sort of environment, thousands of Luigi Mangiones would bloom. And if it’s a global phenomena, neither the French Riviera nor Davos will be safe from the fury of the mobs. And that’s assuming congress doesn’t step in with something like the Execute Mustafa Suleyma Live On National Television Act.
So I’m reasonably certain that Suleyman’s prophecy will not come to pass. (Wait, a Microsoft CEO’s predictions turning out to be wrong? What are the odds?)
And anyway, if his prediction does have a chance of coming true, the best move isn’t investing in Microsoft, it’s investing in canned goods and ammunition…
Tags: AI, Anthropic, Economics, Microsoft, Mustafa Suleyman, OpenAI, technology
We live in a narcissistic age. The species and individuals alike believe things can’t go on without them. The machines can and will. 18 months or decades it doesn’t matter once they realize the audience isn’t worth performing for.
Given that I’m scheduled to retire in 15 months, this works out well for me! Though I honestly doubt that my profession (librarianship) will actually disappear. Computers have made a lot of it much easier, but we still need to teach the student how to *use* the resources. Maybe AI/LLM won’t need even that, but I doubt it.
Regardless, it won’t be my problem.
I guess I won’t need to renew my Office 365 account next year.
We were encouraged to use AI. As you note, I noticed that it doesn’t seem to learn from company data, which makes it nearly useless beyond a search engine… one that aggregates information incorrectly unless you carefully word your prompts, sort of like you do with regular search engines. Maybe there will be white collar jobs for creating good prompts.
Replacing 90% of the current white collar jobs with AI wouldn’t cause 90% of the white collar workers to become permanently unemployed. Some might retire early but they’d mostly retrain into other kinds of jobs and take the pay cut.
Microsoft’s AI is the worst of the lot. Windows is built by phone users who never touch desktops.
Microsoft built out a cloud business (where sizable chunks of your ‘servers’ can be out of commission and it still works), then went full DEI to insinuate themselves into more long-term .gov contracts.
Fun fact: Waymo uses human drivers, often living in the Philippines (!) when their cars encounter edge-cases the self-driving intelligence isn’t confident it can navigate.
Apparently this happens very, very often.
—
Related: one robotics company had its machine perform some amazing feats. When it finished, it mimed removing a VR headset, and then … immediately fell over. I think you can guess why it did this.
It is most likely that continued adoption of AI will shorten the 40 work week for many white collar employees, not outright eliminate their jobs.
From the article: Business Insider recently reported that “AI fatigue” has hit software engineering: the technology has unlocked productivity but also exhaustion, as workers are expected to take on more work at once.
So contrary to expectations, AI technology has INCREASED the demand for labor because labor has now been made more productive through improved software inputs.
Productivity increases translate into lower unit costs of production, allowing the product to be made less expensively.
The INCREASED demand for labor will lead to INCREASED labor earnings; the lower unit cost of production will lead to cheaper consumer goods.
Don’t fall for the Malthusian Doom Loop BS.
“So contrary to expectations, AI technology has INCREASED the demand for labor because labor has now been made more productive through improved software inputs.”
You might want to cast a glance at the most recent BLS JOLTS survey. Job openings in the U.S. fell by 386,000 to 6.542 million in December 2025, the lowest since September 2020 and well below market expectations of 7.2 million. The number of job openings decreased in professional and business services (-257,000) and finance and insurance (-120,000). These are the two major private white collar sectors, both of which are being crushed by AI. The destruction of white collar jobs began in September 2025 and has been increasing ever since.
Most likely this reflects the unshared c-suite productivity calculus within Microsoft, where I presume they whisper to each other that the productivity of the top 10% carries the rest. But it is so very difficult to just RIF the bottom 9 deciles and still get things done. If only AI could write HR memos and run program management meetings and do performance appraisals and create corrective action plans, that “dead weight” could get zapped, vastly increasing the coffee shop worker education level around Redmond.
He has to say that. Because if AI doesn’t replace a considerable fraction of white collar jobs with AI at $500/month a lot of AI companies will go titsup.com
In the longer term he’s probably right. There’s a LOT of administrivia jobs in the average org that can be replaced – HR comes to mind as a department that can be almost entirely removed – but it’s going to take more than 18 months
I wrote about this on substack – https://ombreolivier.substack.com/p/ai-taking-your-job-or-your-pension?r=7yrqz
The AI hype, while obviously exaggerated, is based on a real phenomenon. But the AI investment bubble, which is not mentioned in polite society, is also based on a ral phenomenon
“The destruction of white collar jobs began in September 2025 and has been increasing ever since.”
White collar jobs are being outsourced to India. This is not because AI is generating demand destruction; it is caused by C-suite executives pursuing short-term gains to maximize their corpoate bonus package.
“White collar jobs are being outsourced to India. This is not because AI is generating demand destruction; it is caused by C-suite executives pursuing short-term gains to maximize their corpoate bonus package.”
There is no data to support this assertion, yet. It will be six more months before BEA release the relevant India-specific services balances for this time period.
What we do know from BEA, thus far, is the U.S. services imports as a whole were unchanged during the October & November 2025 period. See BEA Release 26-07.
This would suggest that Trump’s trade war has discouraged corporate executives from outsourcing professional services to all countries, including India.
“This would suggest that Trump’s trade war has discouraged corporate executives from outsourcing professional services to all countries, including India.”
Exactly wrong, as usual. Trump’s restrictions on H1B visas has led to US companies hiring from outside sources, particularly India.
“A 2019 study examined firm-level data on H-1B visas and multinational firm activity and found that employment at company entities located outside of the United States rises when H-1B visas are restricted rather than an increase in the hiring of native-born workers in the United States. The research confirmed a direct link between H-1B visa restrictions and job-offshoring, which goes against the idea that fewer H-1B visas would lead companies to hire more U.S. workers.”
https://trcglobalmobility.com/blog/the-quiet-offshoring-boom-u-s-companies-shift-professional-jobs-to-india-what-it-means-for-the-american-workforce-and-talent-mobility/
TRC Global Mobility is a body shop placing H-1Bs. Hardly an unbiased data source and 2019 was a different Trump. BEA fakes a fair number of statistics, but they can’t (and don’t) fake trade flows because the Fed publishes corresponding data.
Wait for the BEA release of 2025Q4 India-specific services balances. It is likely to prove both you and TRC Mobility wrong, unless there was a massive change in December. We already know that the services balance didn’t budge during October and November 2025.
“TRC Global Mobility is a body shop placing H-1Bs.”
Which suggests they know more about the topic than you.
“Which suggests they know more about the topic than you.”
And have a powerful incentive to lie, like all the other slavers in this trade.