Technology AI Business Insights

What Comes After AI: 5 Moves Already Changing Business in 2026

April 2026

Author

Rodrigo Dantas

7 articles published

I’m going to be straight with you.

Global AI spending will reach $2.60 trillion in 2026, a 47% jump over last year (Gartner, May 2026). In the same window, 56% of CEOs report neither higher revenue nor lower costs from AI over the past twelve months — from a survey of 4,454 chief executives across 95 countries (PwC, 29th Global CEO Survey).

That gap is the story. Not whether AI works — it does. What comes after AI is the reorganisation those two numbers are already forcing: how companies generate value, how much cognitive load people can carry, who controls the electricity, and why human presence is becoming expensive again.

Five moves. All already underway. None of them is a forecast.

1. The $2.6 trillion paradox

Look at where the money actually goes, because this is the part almost nobody reads. Of the $2.60 trillion Gartner forecasts for 2026, $1.43 trillion is infrastructure — chips, data centres, the physical layer. AI services take $586 billion and AI software $453 billion.

So the headline number is not “companies spent $2.6 trillion on AI projects and got nothing.” Most of it never was a project. It was construction. Gartner’s own analyst is blunt about who has been paying:

“Up to this point, AI spending has primarily been driven by technology companies and hyperscalers. Enterprises have yet to really flex their spending potential. Currently, organizations show limited appetite for using AI to drive disruptive enterprise change. Instead, they favor tactical AI initiatives with incremental improvements in efficiency and productivity.”

John-David Lovelock, Distinguished VP Analyst, Gartner (May 2026)

Now the enterprise side. PwC’s 56% is the blunt version; the breakdown is sharper. Thirty percent of CEOs saw revenue rise, 26% saw costs fall, and only one in eight saw both. And MIT’s NANDA initiative went further: across 2025, 95% of organisations reported zero return on generative AI despite $30–40 billion invested (MIT NANDA, July 2025).

Stop for a second and let that number sit. Ninety-five percent.

It doesn’t mean AI is a fraud. It means most companies bought a tool believing they were buying a strategy. They adopted because a competitor did, because the board asked, because the press pushed. Not because they knew which problem they were solving.

Gartner was honest enough to put generative AI in the Trough of Disillusionment for 2026 (Hype Cycle for Artificial Intelligence). Worth noting what that same research says about agentic AI, which got its own Hype Cycle in April: it is still at the Peak of Inflated Expectations, heading toward its own trough. Two technologies, two phases, one conversation that keeps confusing them.

The uncomfortable lesson: if you’re spending on AI without having rethought how your company works, you’re not investing. You’re transferring money to software vendors.

2. AI brain fry: when productivity eats the person producing

This one I write with personal discomfort, because I’ve felt what the data describes.

Boston Consulting Group surveyed 1,500 US workers at large companies and named the thing: AI brain fry — mental fatigue from using, interacting with and overseeing AI tools beyond your cognitive capacity (BCG, March 2026).

Workers with high AI-oversight demands reported 12% more mental fatigue and 19% more information overload than those with low oversight. Fourteen percent said they had experienced brain fry outright. In marketing, 26%.

But the numbers that should reach a P&L conversation are further down the same study. Among workers with brain fry, decision fatigue was 33% higher, minor errors rose 11%, major errors rose 39%, and intent to quit climbed from 25% to 34%.

Read that again as a manager, not as a reader. More serious mistakes and a third of your team thinking about leaving. Turnover, rework and quality loss are three margin destroyers, and not one of them shows up on a productivity dashboard.

Notice where the strain comes from. It is not AI use — it is AI oversight. Monitoring, evaluating and correcting machine output is its own job, and most companies added it to people’s workload without removing anything.

I see it in teams around me. Sharp, capable people who suddenly cannot close a piece of copy, a campaign, a decision, because the tool offers fifteen paths and none feels good enough. The machine is fast. The human mind has a limit. Ignore the limit and what arrives next quarter is turnover and rework.

The most valuable skill in the coming years is not mastering the tool. It’s knowing when to close it and trust your own judgment.

3. The war no one sees: whoever controls energy controls everything

This is the part that shows up least on social media and matters most at the decision table.

Global data centre electricity consumption passes 1,000 terawatt-hours in 2026 — roughly the entire electricity consumption of Japan, a country of 125 million people (International Energy Agency).

A note on a figure you will see quoted everywhere, including in earlier versions of this article: that a single ChatGPT query uses ten times the energy of a Google search. It traces back to a 2024 estimate compared against a Google figure from 2009, and the IEA’s own analysis cuts against it — if every conventional search were run as a simple AI text query, it would add under 4 TWh a year, less than 1% of today’s data centre load. The aggregate demand is real. The per-query comparison is not load-bearing, and I’d rather drop it than repeat it.

What is happening behind the curtain changes the board entirely: the world’s largest technology companies are quietly becoming energy companies.

Microsoft signed a 20-year agreement with Constellation to restart Three Mile Island Unit 1 — 835 MW, targeted for 2028, with 100% of the output going to Microsoft.
Google committed to buying power from seven small modular reactors built by Kairos Power, around 500 MW starting at the end of the decade — the first time a technology company commissioned new nuclear plants outright.
Amazon anchored a $500 million raise in reactor developer X-energy, and separately bought a 960 MW nuclear-adjacent data centre campus from Talen Energy for $650 million.

None of that is a press release about sustainability. It is vertical integration of the one input the whole industry depends on. Whoever secures stable, low-carbon power controls the infrastructure. Whoever controls the infrastructure writes the rules.

4. The in-person economy: presence became the expensive product

In May 2023 the US Surgeon General, Dr. Vivek Murthy, issued an advisory declaring an epidemic of loneliness and isolation — a public health document, not a think piece. It puts stress-related absenteeism attributable to loneliness at roughly $154 billion a year for US employers, plus $6.7 billion in excess Medicare spending.

People are more connected and lonelier than ever, and the market noticed before the models did. Funds now exist for the express purpose of financing nightlife, live events and real gatherings — Best Nights VC positions itself explicitly against that epidemic. I won’t put a growth percentage on the category, because I could not find one that traces to a real source, and a number I can’t verify is worth less to you than the observation itself.

The message for anyone running a business is clear enough without it. The bar became a premium product. The physical store became a differentiator. Face-to-face consulting became a competitive advantage. Everything in person was revalued precisely because it became scarce. In a world where anything can be digital, the analog became luxury. And luxury has margin.

If your company treats human presence as an operating cost, rethink it.

5. The bubble will burst. And honestly, that’s the best thing that could happen.

I know that sentence is irritating. Stay with me a little longer.

MIT Sloan Management Review’s Thomas H. Davenport and Randy Bean named the deflation of the AI bubble as one of five trends for 2026, and the similarities they list are almost embarrassing: sky-high valuations at companies without revenue, growth prized over profit, expensive infrastructure built ahead of demand, and a press that turns every funding round into a civilisational event.

But they add the part that matters most, and that almost everyone quoting them leaves out: this bubble has more in common with the dot-com bubble — which financed the internet backbone — than with the housing bubble. One left infrastructure behind. The other left debt.

The failure numbers support the reckoning. Gartner expects over 40% of agentic AI projects to be cancelled by the end of 2027, and estimates only about 130 of the thousands of self-described agentic vendors are real. Forrester expects three out of four companies that build agentic architectures on their own to fail. S&P Global found that 42% of companies abandoned most of their AI initiatives in 2025, up from 17% the year before. We unpack why that cancellation rate happens, and which processes survive it, in our full guide to agentic AI for business.

“Most agentic AI propositions lack significant value or return on investment (ROI), as current models don’t have the maturity and agency to autonomously achieve complex business goals or follow nuanced instructions over time. Many use cases positioned as agentic today don’t require agentic implementations.”

Anushree Verma, Senior Director Analyst, Gartner

Read that last sentence twice. A large share of what is being cancelled was never an agentic problem — it was an ordinary automation problem wearing a more expensive label.

The point that really matters is different. The bubble is not the end. It’s the cleanup. After the dot-com burst came Amazon, Google, Netflix, Spotify. What actually works tends to be born on the other side of the disaster, because the bubble burns off the noise and what remains is cheaper, more honest and finally accountable to results.

Stanford’s AI Index 2026 — 423 pages of primary data — marks the same turn: from an era of AI evangelism to an era of AI evaluation. Its own findings show why the turn was overdue. Transparency scores are falling. Hallucination rates on user-framed statements run as high as 94%. Benchmark performance still fails to predict real-world results.

The question that until yesterday was “can AI do this?” is now “how well, at what cost, and for whom?” That’s the question separating a serious company from a PowerPoint company.

So the guidance is simple and anything but easy: don’t stop investing in AI. Stop investing without criteria. Define what success is before signing a contract. Redesign the process before automating it. And have the courage to kill the project that isn’t delivering — because the competitors who make it through will come out lighter, faster and more dangerous.

Where value is migrating

If I could reduce all of it to one idea, it would be this: AI will become invisible. Like electricity. Like the internet. Like GPS. It will be in everything and no one will talk about it, because the value was never in the technology. It was in what the technology frees people to do.

And what AI frees up, used with awareness, is room for what no machine replicates: judgment with context, trust built over years, and the presence of someone who is genuinely there — by choice, not obligation.

Four forces are reorganising the value map.

Energy became a first-tier strategic resource. Whoever secures clean, stable power controls the infrastructure, and whoever controls the infrastructure writes the rules.

Process is worth more than tool. The companies extracting real value are the ones that changed how they work first. The ones that layered software onto a broken process will keep breaking, only faster and more expensively.

Human presence became a scarce asset. Every business delivering genuine connection and real attention will appreciate in a way few financial models are pricing today.

Clarity of purpose is a brutal advantage. In a market spending $2.6 trillion against 95% zero return, the company that knows exactly what it is doing, and why, holds something no algorithm can buy.

I have no idea what the next hype will be, and I’m suspicious of anyone who says they do. But I know what comes after AI in the sense that matters: consequence. And consequence is the only territory where real businesses get built.

What I feel, looking at all of it together, is that we’re living the end of a cycle of enchantment and the beginning of a cycle of responsibility. Less exciting, I know. Infinitely more real.

The bill has come due. I’d rather be on the side of those who read it first.