AI is here. Now what?

AI has become a fact of business life, or will be shortly.
We are all trying to work out where AI fits into what we do.

The answer starts with the two directions AI already takes:

Through people

AI enters the business from both sides of the payroll. Employers often find their staff using personal AI chat accounts before the company pays for its first licence. The know-how is already on the payroll. Seen from the top of an organisation, AI in the team’s hands works both ways. Put to clear tasks, it frees hours and lifts the quality of the team’s work. Used loosely, it burns money on usage fees and pulls people away from the jobs they were hired for. For the employee, AI does the heavy work while their own experience and judgement supply the context.

Through software

AI reaches a company on terms companies already know from the cloud: rented, and paid for by use. They moved their servers to Amazon, Microsoft and Google and now pay for computing as a service, by what they use. AI as software follows the same path: knowing how it works inside is the job of whoever builds and runs it. AI arrives inside programs the company already pays for: Copilot in Word and Outlook, Gemini in Gmail and Docs, call summaries in Teams and Zoom, generative fill in Photoshop. In the last week of September 2026, Microsoft took this furthest: its redesigned Copilot brings Word, Excel and PowerPoint into one app and adds an AI agent that works across a company’s meetings, documents and emails.^{1 2} Beyond office tools, agents and chains of agents take over whole processes inside the company’s systems, and people review the results. The strategic call stays with the company: which process AI takes over, what that process must return, and when AI has earned its place.

For now, both point the same way:

AI gets added to what is there.
People use AI inside their current work, and companies attach AI to the systems they run.

The reason sits in the systems themselves: they carry rules, exceptions and history collected over decades, and rewriting them costs too much and risks too much.³ WordPress, which runs about 41% of all websites, did exactly that in May 2026 and connected AI models to the core of its platform.^{4 5}

For whoever is in charge, both come down to the same questions:

What AI speeds up, what it improves, and how it moves costs and revenue.

Where AI goes next is still open. For people, working with AI is becoming part of the job description. In the company’s systems, agents take over more of each process, preferably in stages, each one proving its value first. As long as rewriting old systems costs more than connecting AI to them, AI will keep arriving as an addition. Companies built around AI from their first day are already testing the other route.