This year’s AI reality check is in the mail — again

Welcome to 2026, the year when generative AI and AI agents will demonstrate real business value, go from hype to reality, from pilot to production. At least that’s what I’ve heard, and I’m guessing you’ve heard the same thing. The precarious thing about it is that it sounded exactly the same, in exactly the same words, a year ago. 2025 was the year when all of this was going to happen.

Of course, that wasn’t the case. Absolutely, the technology continued to make strides and the biggest thing over the past year was the launch of coding agents, led by Claude Code, which quickly became the extended arm of many developers. Coding is still the clearest use case for generative AI in business.

But otherwise, AI in the enterprise seems to be going the same way as it was in 2024. Almost everyone is trying, but few are succeeding. A recent example: According to a CFO survey from Duke University and two regional central banks, released in December, a majority of US CFOs say they have not seen any impact from AI on productivity, decision-making, customer satisfaction, or time spent on high-value tasks. For 2026, a greater impact is expected, but only in single-digit percentages, surveyed CFOs said.

This resonates with reports about AI suppliers such as Microsoft, who are still having difficulty convincing companies of the excellence of Copilot and similar AI assistants with diffuse utility.

At the same time, cooler winds are blowing among the general public, whose attitude should never be underestimated. Sure, people love their ChatGPT and asking it things they shouldn’t ask it about, but in parallel with this, “slop” became the word of the year in English in 2025 and the first “AI chatterbox” of 2026 was Elon Musk’s porn bot Grok, which undresses both adults and children on X.

It was also quite refreshing to see analyst firm Forrester last week come out to puncture the narrative that AI is already causing mass layoffs. In the firm’s view, the large waves of layoffs, especially in the US, which are said to be due to AI, are actually entirely ordinary financially motivated cutbacks — but are conveniently blamed on AI. Forrester has therefore scaled back its “AI takes the jobs” forecast. Just 6% of US jobs are threatened by 2030, the firm predicts — not insignificant but certainly a completely different magnitude than many other forecasts.

Forrester’s research can also be kept in mind when reading news that Swedish IT students from polytechnics are having a much harder time getting jobs today, which is directly linked to AI. And sure, maybe the need for the most junior developers has decreased when Claude Code can do more of the simple jobs. But as MYH also says, the main reason is the recession. It’s hard for anyone to get a job today, not least in tech. And the trend is not new either; MYH said the same thing — although without mentioning AI — a year ago.

It’s no wonder that new technological advances take a long time to take hold, or that it’s always difficult to separate hype from substance. But it’s important to remember, not least for business leaders, that the responsibility lies with the technology suppliers and no one else. They’re the ones to point the finger at when technology doesn’t live up to what was promised, not the IT managers who do the best they can with the tools they offer.

Without even going into the “AI bubble,” which is related but partly a separate phenomenon, one can probably predict that generative AI and its agent force are facing a proper reality check in 2026. Whether that reality check results in 2026 actually being the year when AI shows business benefit, well, that remains to be seen.

Otherwise, we might hear about it again in 2027. As we supporters of Swedish football club AIK usually say: “Next year, then f*ck it!”