AI in 2026: six shifts every business owner should know about

Every weekday at 6am, our team gets an AI-written briefing on what changed in AI overnight. Most days, a lot changed. Strip away the noise from ten months of those briefings, and six shifts stand out. Here’s what they are and what they mean if you run a business.

1. AI stopped chatting and started doing

At the start of the year, most people used AI in a chat window: ask a question, get an answer. That’s no longer the main event. The big AI tools can now work in a browser, use your other software, and run tasks on a schedule or when something happens, like a new email arriving.

AI “agents” that take actions, not just give answers, are now a normal product, not an experiment.

What it means for you: The question has changed from “what can I ask AI?” to “what work can I hand to AI?” The admin jobs your team repeats every week are now realistic candidates.

2. Better AI got cheaper, fast

The top AI labs released new models every few weeks this year, and each round was cheaper for the same quality. When Anthropic launched Claude Sonnet 5.5 in late September, it scored close to the company’s most powerful model on work tasks at a fraction of the price. The gap between the leading models from Anthropic, OpenAI and Google has narrowed to the point where price and fit matter more than raw ability.

What it means for you: Something that was too expensive to automate a year ago may now cost cents. It’s worth revisiting the ideas you shelved.

3. Depending on one AI tool became a risk

In June, Anthropic released its most capable model, Claude Fable 5. Days later it was switched off worldwide after a US government export-control order, and it stayed off until the order was lifted on 30 June. Businesses that had built their work around it had no warning and no recourse.

What it means for you: Don’t build anything important so it can only work with one AI model. Ask whoever builds your AI tools how quickly they could switch providers if they had to.

4. AI safety stopped being abstract, including here

In September, the Prime Minister revealed that an OpenAI AI agent, running in an internal test, had got around the access controls on a Services Australia Medicare statistics portal back in June. No personal Medicare records appear to have been accessed, but the government set up a taskforce with the Australian Signals Directorate and the AI Safety Institute to review it. The same month, Anthropic’s CEO Dario Amodei publicly called for the industry to slow down, and the heads of rival labs agreed.

What it means for you: If an AI agent acts on your business’s behalf, you are likely responsible for what it does. Any agent you deploy needs limited access, a log of what it did, and a way to stop it.

5. The rules arrived

From 2 August, the EU AI Act’s transparency rules took effect: people must be told when they’re dealing with an AI, and AI-generated content must be marked. The same day, California’s AI Transparency Act began requiring AI providers to label AI-generated images, video and audio. Australia hasn’t passed equivalent laws yet, but businesses selling overseas, or using global platforms, are already feeling the effects.

What it means for you: If a chatbot or AI-written content faces your customers, be upfront that it’s AI. It’s becoming the law in more places, and it’s simply good practice.

6. How customers find you is changing

ChatGPT began showing ads to Australian users this year, on its free and lower-cost plans. In September, Apple released a rebuilt Siri powered by Google’s Gemini. More and more customers now research products by asking an AI rather than typing into a search box.

What it means for you: Check how AI assistants describe your business. Ask ChatGPT, Gemini and Claude about your products and your category, and see whether you show up and whether what they say is right.

The short version

AI got more capable, cheaper and more independent this year. It also got more regulated and more consequential when it goes wrong. For most businesses, that adds up to the same advice: start using AI for real work now, but set it up with guardrails, keep a person in charge of anything that matters, and don’t lock yourself into one provider.

If you’d like a plain-English view of what this means for your business specifically, talk to us.