AI trends worth planning around in 2026

Every January produces a list of AI predictions, and most of them age badly. This one looks at what has already changed, and at what that means for how a business should plan, using figures from the surveys and analysts we rely on.

1. AI that does the task, not just the answer

The biggest change is from AI that responds to a question to AI that carries out a job. An agent can break a goal into steps, use your business software, check its own work and carry on. Gartner expects 40% of enterprise applications to include task-specific AI agents by the end of 2026, up from less than 5% a year earlier.

The same firm also expects over 40% of agentic AI projects to be cancelled by the end of 2027, mainly because of cost, unclear value and weak risk controls. The planning lesson is to pick one workflow with a clear measure, set limits on what the agent can do, and keep a person responsible for the outcome. We cover this in detail in agentic AI for business leaders.

2. Models that think before they answer

Newer models can spend time working through a problem step by step before replying. That makes them much better at analysis, planning, maths and code than the first wave of chat tools, and it widens what is worth trying: reviewing a contract against a checklist, reconciling figures across reports, drafting a plan from messy notes.

They are slower and cost more per answer, so the sensible approach is to use them for the hard problems and a smaller, quicker model for routine ones. Our guide to choosing a model explains how to test that on your own work.

3. AI inside the software you already use

The most common way people meet AI isn’t a separate product. It’s a button inside the CRM, the email client, the accounting package or the office suite. For a company starting out, that is good news: you can often test an idea using features you already pay for, without a new contract or a new system to learn.

It also changes the buying question. Before you sign for a new tool, check what your current suppliers have added, and what it does with your data.

4. The gap between using AI and getting a return

McKinsey’s State of AI survey found 88% of organisations using AI in at least one function, but only about a third scaling it. Reports on pilot results are harsher: MIT’s NANDA initiative found that roughly 95% of generative AI pilots produced no measurable return, pointing to integration and how organisations work as the causes.

The trend underneath is a shift in attention from “are we using AI?” to “what did it change?”. Companies that can answer with a number, such as hours saved, faster replies or fewer errors, are the ones that will keep funding it. Our article on why AI pilots stall covers the three questions that decide this.

5. Data, security and regulation move to the front

Poor data is the quiet reason many projects fail. Gartner found that 63% of data management leaders either lacked the right data practices for AI or weren’t sure they had them. Security teams are using AI to spot threats faster, and attackers are using it to write more convincing phishing messages, so both sides of the equation are moving.

Regulation is catching up as well. In the EU, the AI Act’s timetable shifted in 2026, but the transparency duties and the obligations for high-risk uses still apply on dates that businesses outside Europe can also fall under. We explain what changed in the EU AI Act Digital Omnibus.

What to do with this

You don’t need to act on all five. A sensible short list:

1. Choose one process that costs real time or causes frequent errors, and measure it today.
2. Look at what your existing software can already do before buying anything.
3. Check your data for that process: where it lives, who owns it and whether it is consistent.
4. Write a short policy on approved tools and what information can go into them.
5. Review progress every month against the baseline, and stop what isn’t working.

If you’d like a view of where you stand first, the AI Readiness Index scores strategy, data, systems, people and governance in about six minutes. And if you’d like to talk through which of these trends is worth your attention, you can book a session with us.