Short answer: In its third-quarter 2026 results, published on October 28, 2026, UBS said employees run more than 12,700 personal AI agents for daily tasks (UBS). The bank has 608 live AI use cases in production, up 81% year over year, and more than 1,090 further applications in development. More than 18,000 of its software developers use AI tools.
Key Takeaways
- 12,700+ personal AI agents used by UBS employees for daily work.
- 608 live AI use cases in production, an 81% increase in 1 year.
- 1,090+ applications in development.
- 18,000+ developers use AI tools in the software development cycle.
- 9 large-scale, end-to-end AI programs run alongside the personal agents.
- 30 million+ insights generated year to date by STAAT, a tool for financial advisor teams.
What did UBS report?
The AI figures sit inside a strong quarter. UBS reported net profit of $2.7 billion, up 8.4% year over year, and profit before tax of $3.6 billion, up 27%. Invested assets reached $7.3 trillion at September 30, 2026 (UBS).
The bank describes 2 layers of AI work. The first is broad access: thousands of employees build and use personal agents for their own tasks. The second is a set of 9 large-scale, end-to-end transformation initiatives that change whole processes. UBS also ranked as the leading bank in Europe in the 2026 Evident AI Index, and the only European bank in the global top 10.
How do the numbers compare?
The 608 live use cases are the more useful figure for other enterprises, because they count work in production rather than experiments. The 81% annual growth shows a pipeline that converts: the 1,090 applications in development are almost twice the number already live.
Other banks report similar momentum. Customers Bank closed select commercial loans in 7 days in an AI pilot, against a typical 30 to 60 days, and its parent doubled its deployed AI agents in the most recent quarter (PYMNTS). A PYMNTS Intelligence survey found 85% of financial services firms plan to raise AI budgets over the next 12 months, with adoption concentrated in structured back-office tasks such as accounting close and credit risk scoring.
What about headcount?
UBS internal and external personnel fell to 109,990 at September 30, 2026, from 122,382 a year earlier, a drop of about 10%. The decline coincides with the Credit Suisse integration, which has delivered $13.2 billion in cumulative gross cost savings. About 90% of in-scope legacy Credit Suisse applications are now fully decommissioned. UBS does not attribute the headcount decline to AI, so the 2 trends should not be read as cause and effect.
What does 12,700 agents mean for governance?
Each personal agent acts with some level of access to data and systems. At this scale, ownership, identity and credential management become operational questions. IBM cited a Gartner projection on October 27, 2026, that the average Fortune 500 company will run more than 150,000 agents by 2028, and that only 13% of organizations are confident they can govern them (IBM). Banks face the added layer of regulatory review for any agent that touches client data or advice.
Related: IBM launches identity and control for AI agents and securing enterprise AI agents.
What should enterprises do now?
- Run 2 tracks. Give employees governed access to personal agents, and fund a short list of end-to-end process programs.
- Measure production, not pilots. Track live use cases and conversion from development, as UBS does.
- Give every agent an owner. Register agents, assign accountable owners and use short-lived credentials.
- Plan the talent mix. Scaling agents needs AI engineers, platform and security engineers, and product owners who understand the business process.
- Start with structured work. Back-office tasks with clear inputs and audit trails show results first.
Related: why AI agent projects fail and agentic automation in banking.
Frequently Asked Questions
How many AI agents does UBS use?
UBS said in its third-quarter 2026 results, published on October 28, 2026, that employees run more than 12,700 personal AI agents for daily tasks.
How many AI use cases does UBS have in production?
UBS reported 608 live AI use cases in production, up 81% year over year, with more than 1,090 further applications in development.
How many UBS developers use AI coding tools?
More than 18,000 UBS software developers use AI tools in the development cycle, according to the bank.
Did UBS headcount fall while AI use grew?
Yes. Internal and external personnel fell to 109,990 at September 30, 2026, from 122,382 a year earlier. The decline coincides with the Credit Suisse integration, and UBS does not attribute it to AI.
What can other enterprises learn from UBS?
Pair broad personal agent access with a small number of large end-to-end programs, track live use cases rather than pilots, and build governance for agent identity and access before agent counts grow.
Hiring AI engineers or scaling an agent program? Talk to Deployed Labs.






