AI That Works in the Background: October 2026's Big Shift
For the last few years, using AI has meant opening a chat window, typing a question and waiting for an answer. In the space of six weeks this autumn, the biggest AI companies all moved in the same direction: AI that works on its own in the background, and only comes to you when something needs a decision.
What launched
- Meta Muse arrived on 8 September, followed by OpenAI's Dots on 29 September, both pitched as always-on personal agents.
- On 16 October, OpenAI added parallel agents to GPT-6 Pro in ChatGPT, splitting a single task into parts that several agents work on at the same time.
- On 21 October, Anthropic announced Claude Duet as a research preview. It's described as a "chief of staff" for work: it sends a morning brief across your connected apps, keeps working through its queue during the day, and asks for your approval before taking any major action.
Most of these are early. Claude Duet, for example, is currently a waitlist for paid subscribers in the US. But when every major lab ships the same idea within a few weeks of each other, it's a clear signal of where things are heading.
Why this matters for a small business
The shift is from AI you have to remember to use, to AI that does the routine checking for you. Think of the jobs that eat a small business owner's week: chasing unpaid invoices, spotting the enquiry that needs a reply today, pulling together the numbers for a Monday catch-up. These are exactly the tasks always-on agents are designed for.
The approval step is the important part. None of these tools are designed to run your business unsupervised. They're designed to do the legwork and hand you the decision, which is the right balance for anything involving money, customers or legal commitments.
What it doesn't change
An agent is only as useful as the systems it can see. If your enquiries live in three inboxes, your invoices in a spreadsheet and your customer list in someone's head, a background agent has nothing reliable to work with. The businesses that will get value from this first are the ones with tidy, connected processes, and that's true whether you use Anthropic, OpenAI, Meta or something built for you.
What to do now
You don't need to sign up for every waitlist. A better first step is knowing which of your processes are worth automating at all, and which tools you already use that an agent could connect to. That's exactly what our AI opportunity audit covers: a ranked list of where AI would save you time or money, and a costed plan to get there. If you already know the one job you want off your plate, the automation starter gets a single workflow running on its own in a few weeks.
Still unsure how far behind or ahead your business is? Our piece on the UK small business AI agent adoption gap is a good place to start.
Keep reading
Why Most Successful AI Projects Have an Outside Partner Involved
Most AI projects that actually deliver value weren't built entirely in-house. Here's what that means if you're a growing business deciding how to start.
AI Just Got a Lot Cheaper. Here's Why That Actually Matters.
OpenAI's newest model offers near-flagship performance at a fifth of the price. For a business that's been waiting for AI to be worth the cost, that's the real story.