UBS, one of the largest wealth managers in the world, published its third-quarter 2026 results on 28 October. In the AI section of that update, the bank wrote that its employees are running over 12,700 AI personal agents to support them in their daily activities. It is one of the clearest numbers any large company has put on how far personal AI agents have spread inside its workforce.
What UBS disclosed
The figures below come directly from UBS's quarterly reporting page:
- Over 12,700 AI personal agents run by employees.
- 608 AI use cases live in production, an 81% increase on a year earlier.
- More than 1,090 further applications in development.
- Over 18,000 software developers using AI tools in their development work, which UBS says has led to measurable productivity gains.
- Nine large, end-to-end transformational AI initiatives in progress.
Some coverage of the results gave the release date as 29 October. UBS's own page says the results were published on 28 October.
What UBS did not say
The update does not name the platform the agents run on, and it does not say whether employees built them themselves or picked them from a central catalogue. Any claim about which vendor powers them is therefore a guess for now. The update also does not define what counts as a personal agent, so the 12,700 figure is best read as a sign of scale.
The headcount number next to it
The same results show UBS counted 109,990 internal and external staff on 30 September 2026, against 122,382 a year earlier. That is a drop of 12,392 people. UBS does not connect that change to AI in its update, and a bank of that size has many reasons for headcount to move, so it would be a stretch to read the two numbers as cause and effect. The two figures do sit side by side in the same report, and they say something about the kind of efficiency large employers are now chasing.
Why personal agents matter
For the last few years, most corporate AI announcements were about central projects run by IT or innovation teams. The word personal is the interesting part here. It suggests agents are becoming something each employee runs for their own tasks, much like a spreadsheet or an inbox rule. Our honest view is that this is where the real change in daily work happens, because it is decided task by task by the people doing the job.
For anyone working in a large company, the practical meaning is that the skills that matter now are describing a task clearly, checking the result and knowing when not to delegate. Those skills are easy to practise now on small, low-risk jobs, long before an agent is handed anything that touches clients, money or legal commitments. The people who learn that early tend to be the ones asked to set up the team's agents later.
What to do this week
Write down three tasks you repeat every week, such as a status update, inbox sorting or meeting preparation, and decide which one you would trust an agent with first and who would check its work. Then ask your manager the same question for your team. If you want to try an agent yourself, our tool stacks show what to start with, and our earlier piece on how experts work with AI agents covers how to keep control of the output.