AI hasn’t taken your job. But it’s already taking plenty of money

Sam Altman admitted he was wrong. The visionary who spent years predicting inevitable mass unemployment driven by artificial intelligence recently made a surprising statement at a Commonwealth Bank of Australia conference. He is "delighted to be wrong" about how quickly AI would eliminate jobs. Altman is now pushing back against the white-collar apocalypse that most Silicon Valley experts – including himself – once forecast.

It’s worth asking why the CEO of OpenAI suddenly turned so "pessimistic" about his own industry. The AI hype shows no signs of cooling, and plenty of office workers have already convinced themselves the technology is about to render them obsolete.

A quick terminology check

When media, investors, and managers talk about "artificial intelligence," they almost always mean Large Language Models, or LLMs. This is not intelligence in any meaningful sense – it’s a statistical engine that predicts the next word in a sequence based on trillions of examples scraped from the internet.

ChatGPT, Claude, Gemini – all LLMs. They don’t think, don’t understand, and certainly don’t feel. They generate plausible text. They do it well. And they do it at enormous cost.

True artificial intelligence – capable of learning, setting goals, and acting autonomously in an open-ended environment – remains either science fiction or a very distant prospect. The gap between "very smart autocomplete" and genuine reasoning is vast. AI startup founders tend to avoid this topic, since blurring the distinction makes it considerably easier to raise billion-dollar rounds.

Business is starting to sober up

Reality is beginning to bite. Chaac Pizza Northeast, one of Pizza Hut’s largest franchisees with over 110 locations on the US East Coast, is suing parent company Yum! Brands for $100 million. The franchisee was forced to adopt an AI dispatch system called Dragontail. Instead of speeding up deliveries, it slowed them down. On-time delivery rates dropped from 90% to less than half, and New York sales fell by roughly 10%. "Dragontail did the exact opposite of what it promised," the lawsuit states.

Starbucks, meanwhile, quietly pulled an AI inventory tool from all its North American stores just nine months after launch. The system couldn’t reliably count bottles of syrup. In the promotional video from the launch event, the scanner visibly missed a bottle of peppermint syrup. An omen, in hindsight.

Uber’s COO has publicly acknowledged that AI spending is increasingly hard to justify. "If you’re not actually able to draw a direct line between token usage and useful features shipped to users, that trade becomes harder to justify," he said — after the company burned through its entire annual AI budget in four months.

Microsoft is counting the cost too. In May 2026, the company began revoking Claude Code licenses for engineers across its Experiences & Devices division — the teams that build Windows, Teams, and Surface. The reason was straightforward: token-based billing had exhausted the annual budget well before the year was out. 

A company that has invested roughly $13 billion in OpenAI and aggressively champions AI across its product lines couldn’t survive its own bill. 

The scale of "optimism"

Research firm Gartner warned back in 2025 that more than 40% of agentic AI projects would be cancelled by the end of 2027, due to escalating costs, unclear business value, and weak returns on investment.

Despite this, Nvidia CEO Jensen Huang used GTC 2026 to propose giving engineers a token budget worth half their annual salary to drive more productive AI use. Under his framework, an engineer earning $500,000 a year should be spending another $250,000 on tokens. "Otherwise I will be deeply alarmed," Huang said.

The industry, meanwhile, keeps moving on momentum. According to Gartner’s latest forecast, global IT and AI spending in 2026 will exceed $6.3 trillion, with the AI sector alone accounting for more than $2.5 trillion. These astronomical figures keep climbing – with no guarantees of a return.

So what comes next?

I don’t believe in the apocalypse. But I don’t believe in magic wands either.

AI is a powerful and expensive tool currently passing through what technologists call the "trough of disillusionment" – the phase that follows every peak of inflated expectations. We’ve seen this before: the dot-com crash of the early 2000s, the blockchain craze of 2018 that was going to "replace the banking system any day now."

Altman walking back his predictions, Pizza Hut filing lawsuits, Starbucks returning to hand-counting syrups – none of this is a collapse of the technology. It’s the normal process of a technology growing up.

Dismissing AI entirely would be foolish. LLMs are already delivering real, measurable value to doctors, lawyers, journalists, and engineers. That value is genuine, and the tools arriving in the next few years will be more capable still.

But panic and blind faith are equally unhelpful. While AI hasn’t taken your job, it is reliably taking enormous sums of money from everyone who bought the hype in bulk.

The difference between "artificial intelligence" and a "large language model" is the difference between a mind and a very clever parrot. Both have their uses. But nobody is handing the parrot the keys to the company just yet.

At least not right now.

/Business/AI hasn’t taken...