How the AI transformation will play out in 2026

Buckle up, people, the ride is just getting started

Want to know the future, dear reader?

Of course you do.

I’ve communed with the AI platform gods – ChatGPT, Claude, Deep Seek, Perplexity and Grok, in this instance. They provided the following forecasts for how AI will embed itself even deeper into your life over the next 12 months.

It will be the year of the agent
When tech bros bang on about ‘AI agents’ or ‘agentic AI’, they’re not referring to slightly smarter chatbots. They’re pointing to near-magical tech that can move through your digital life largely unassisted: reading emails, updating the CRM, filing tickets, poking the finance system and booking meetings.

Instead of “help me write this reply”, the brief becomes closer to “take every overdue invoice from this list, chase it, update the records and flag anything dubious‑looking for me to review”.

The human sets the intent and the guardrails; the machine trudges through the to‑do list.

This marks the shift from AI being a ‘copilot’ to a colleague. Possibly a senior colleague.

Copilots sit in one app, helping one person. Agents roam across tools. A half‑decent 2026 agent will be able to watch a support inbox, generate and send responses, create or close tickets, log what happened, and schedule follow‑ups, with the human stepping in only when something truly out of the ordinary happens.

As with many things related to AI, the automation of many more tasks is good or bad depending on where you happen to be located in the food chain.

Agents are coming for the ‘glue work’ that props up modern white‑collar employment: the status updates, the nudges, the form‑filling, the ‘just circling back’ emails.

Once agents can reliably handle all the glue work – and that day is not far off – one person can supervise much more activity than before. For some workers, it will feel like they’ve gained superpowers. For others, it looks like their job being quietly decomposed into the small set of decisions a machine still needs them for.

AI everywhere, all of the time
The big models keep getting bigger brains and longer attention spans. That matters for agents because it makes them less like overeager interns and more like someone who can actually hold a plan in their head for more than five minutes.

That means you can hand over a messy, multi‑step job and trust the system not to lose the plot halfway through.

But the more interesting shift is at the other end of the spectrum – the small, efficient AI edge models that run on your laptop, phone, car, or office server.

Instead of every interaction bouncing up to a distant data centre, a lot of work – transcription, summarisation, basic coding, routine analysis – happens on the device in front of you. That cuts latency, lowers costs, and, crucially, keeps data where nervous CIOs and regulators want it to stay.

Put those two trends together and you get ‘AI everywhere’ in a literal sense. The heavy artillery sits in the cloud, doing the hard thinking and long-range planning. The lighter models sit at the edge, glued into everyday tools. Workers will no longer ‘use AI’; it will be threaded through all the infrastructure they already rely on.

There will be exciting breakthroughs, but it’s unclear who will benefit
The lab bench and the data centre are where ‘AI-enabled massive productivity gains’ stop being an abstract concept and blossom into new drugs, new materials, and new forms of leverage for whoever owns them.

On the science side, you can already see the outlines.

Models aren’t just summarising papers any more. They’re now helping design experiments, propose new molecules, and drive ‘self‑driving labs’ where robotic systems run 24/7, testing hypotheses humans would never have had time to explore.

That will result in cheaper or more effective batteries, treatments and industrial processes.

As always, the question is who captures the benefits. A handful of industrial conglomerates, or a broader ecosystem of researchers, smaller firms and public institutions?

As welcome as many scientific breakthroughs will be, they’re unlikely to be all upside. They’ll impact labour markets that are already bifurcating into AI haves and have‑nots, and in states that are still arguing over how generous they want to be to displaced workers.

Faster scientific discovery and more productive firms are, in principle, good news. But if the benefits are captured by a handful of companies – and the C-suite and shareholders of those companies – while a growing chunk of the workforce is stuck in low‑wage service sector or insecure gig work, the politics get ugly quickly. 2026 is less about whether AI boosts productivity (it undoubtedly will) and more about whether anyone outside the winners’ circle feels those gains in their pay packet or quality of life.

The era of light-touch regulation is very much over
Earlier this week, Australia became the first country to ban social media for children under 16. Social media isn’t AI, but the takeaway message is clear. With the possible exception of the US, Western governments are no longer inclined to give the tech companies free rein. After belated awakening to the downsides of excessive phone and social media use, politicians are unlikely to let the AI experiment play out for a decade or two before stepping in.

Systems that make decisions about credit, hiring, healthcare or education don’t just have to work; they must also be explainable, auditable and defensible in front of a regulator or a judge.

There’s also a quieter but equally important set of constraints – insurers, auditors and big‑company procurement. Even in places where the law is vague, these actors effectively enforce soft law by refusing to underwrite or buy systems that don’t meet governance standards.

The net effect is a shift in power. A small number of big platforms and well‑capitalised firms can afford to navigate this landscape. Smaller players and individual workers will find themselves living with rules they didn’t write but now have to obey.

And a merry Festivus to you all
And with that, I’m signing off for a month or so.

As always, thanks to all those who’ve read, liked, shared or critiqued my posts this year. Your support and feedback are always much appreciated.

See you on the other side.