🔍 Read the full analysis: A Closer Look At What UBS Doesn’t Say About Its AI Agents on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
UBS reported on Oct. 28 that employees were running more than 12,700 AI personal agents, alongside 608 AI applications in production and more than 1,090 in development. The bank has not described the agents’ permissions, oversight or uses in detail, and says it does not allow outside AI models to act for clients.
UBS said on Oct. 28 that employees were running more than 12,700 AI personal agents to support daily work, putting a large deployment figure in the Swiss bank’s third-quarter results. The release also counted 608 AI applications in production and more than 1,090 in development, but did not explain what the personal agents can access, how their use is monitored or whether they can take actions without human approval.
The figures appeared in a section of UBS’s third-quarter results release about investment in the bank’s business. UBS also reported nine large-scale AI initiatives and said more than 18,000 software developers were using AI tools in development cycles. The bank described the resulting productivity gains as measurable but did not publish a figure for them.
UBS reported that its 608 live AI use cases represented an 81% increase from a year earlier. It listed more than 1,090 additional applications in development. Reuters and PYMNTS reported the 12,700-agent figure, citing the results; neither supplied further detail on what those agents do.
The release also said the bank’s US financial-adviser teams had received more than 30 million insights so far this year from STAAT, a tool that provides suggestions. That figure describes insights generated, not necessarily actions taken or client outcomes. UBS did not link any of these AI figures to its quarterly staffing changes.
Infographic 01 · What UBS reported
12,700 personal AI agents, inside one bank
The AI figures from UBS’s third-quarter results, and how fast they grew in three months.
From permission to 12,700 agents
UBS 2Q26 release (29 Jul 2026) vs. 3Q26 release (28 Oct 2026)
What the agent number is not
- Not 12,700 people. One employee can run several agents; UBS does not say how many people use them.
- Not a definition. UBS does not say what an agent can access or do on its own.
- Not a result. “Measurable productivity gains”, but no hours or costs published.
- Scale, for context: about 13 agents per 100 internal full-time staff (12,700 ÷ 97,593, own calculation).
“Our employees are running over 12,700 AI personal agents to support them in their daily activities.”UBS, third-quarter 2026 results release, 28 October 2026
Also in the release: 9 large-scale AI initiatives and more than 30 million insights from the STAAT tool for US financial advisor teams so far this year. UBS placed 6th of 50 banks in the 2026 Evident AI Index.
Sources: UBS 3Q26 release, 28 Oct 2026 · UBS 2Q26 release, 29 Jul 2026. Reported identically by Reuters, 28 Oct 2026 and PYMNTS (29 Oct 2026).
Powered by Thorsten Meyer AI
The Oversight Challenge Behind 12,700 Agents
The disclosure matters because it shows employee-facing AI is already operating at substantial scale inside a highly regulated bank, even as public information about its safeguards remains limited. For staff, agents could help with routine research, drafting and organization. For the bank, tools that handle internal information raise questions about data access, review and accountability.
The distinction between preparing work and taking action is central. A tool that drafts a summary for an employee differs from one that can send a message, change a record or approve a transaction. UBS’s results do not say which capabilities its personal agents have. Without that information, the agent count alone cannot show the extent of automation, its business impact or the risk involved.
The figures also offer other companies a practical governance question: when employees can create tools themselves, how does an organization keep track of what is in use and what permissions it has? The relevant measure is not just how many agents exist, but what they can do and who checks them.
Products Worth Considering
Apple Mac mini Desktop Computer with M4 chip, 10‑core CPU and GPU, 24GB Unified Memory, 512GB SSD, Gigabit Ethernet
FLEXISPOT 42-Inch Electric Standing Desk Converter
FLEXISPOT 42-Inch Electric Height-Adjustable Desk Converter with Keyboard Tray
As an affiliate, we earn on qualifying purchases.
From Employee Access to Bank Controls
UBS had described a broadening of employee access before the third-quarter announcement. In its July 29 second-quarter release, the bank said all employees had the ability to create their own AI agents. It said this could help staff spend more time on judgment, collaboration and client work. That statement describes access and the bank’s intended benefit; it does not establish how every agent is used or how much time it saves.
UBS’s technology materials describe an internal assistant called Red, rolled out to around 100,000 employees, and an AI-powered HR assistant for tasks such as holiday bookings and policy questions in Microsoft Teams. These examples help show the range of staff-facing tools, but UBS has not said whether either system is among the 12,700 personal agents or defined exactly what it counts as an agent.
The bank has drawn a separate line around AI acting for customers. At an event at ETH Zurich on Oct. 27, CEO Sergio Ermotti said UBS was not allowing outside models into the bank to act on clients’ behalf, including to manage accounts or move deposits, according to Reuters. That restriction concerns external models acting for clients; it does not, by itself, describe the permissions or controls for employee-run agents.
“Employees “are running over 12,700 AI personal agents to support them in their daily activities.””
— UBS, in its third-quarter results release
What UBS Has Not Specified
UBS has not publicly defined what it counts as a personal AI agent, identified the platforms on which they run or disclosed how many employees use them. The results do not explain what internal systems or data an agent may access, whether agents can take actions independently, or how outputs are checked and recorded.
The bank also does not say how many of its 608 production use cases involve customers, nor does it provide a breakdown by risk or business function. Its reported developer productivity gains lack a numerical measure in the release. These gaps mean the figures establish deployment scale, but not the agents’ effectiveness, safety record or effects on client service.
UBS and Reuters linked a roughly 1,500 full-time-equivalent decline in internal staff during the quarter, to 97,593, to the Credit Suisse integration. Neither attributed that reduction to AI. The available material does not establish whether AI affected staffing, in either direction.
Further Detail on Controls and Results
The next useful disclosures would be a clearer definition of the agents, their permitted actions and the controls governing access to sensitive information. UBS could also explain how employees review agent outputs, how incidents are handled and how the bank measures the productivity gains it references. The third-quarter release does not set a date for such disclosures.
Regulatory expectations make inventory and oversight relevant beyond UBS. The source report says Switzerland’s FINMA expects banks to maintain a complete AI inventory with risk classifications and has warned that keeping inventories is harder as applications become more accessible. Whether UBS will provide a public breakdown of its inventory or agent safeguards is not yet clear.
For now, the confirmed development is the reported number of employee-run agents and the bank’s stated boundary against outside models acting for clients. More information would be needed to judge the agents’ actual capabilities, their business results and how consistently they are governed.
Key Questions
What did UBS report about its AI agents?
UBS said in its Oct. 28, 2026, third-quarter results that employees were running more than 12,700 AI personal agents to support daily activities.
What can the UBS personal agents do?
The results do not define their tasks or capabilities. UBS has described employee-created agents and other staff tools, but it has not confirmed which systems or functions make up the 12,700 figure.
Can UBS AI agents act on client accounts?
Ermotti said UBS was not allowing outside AI models to enter the bank and act for clients, such as managing accounts or moving deposits. That statement does not spell out the permissions of employee-run agents.
Did UBS link AI to its staff reduction?
No. UBS and Reuters attributed the roughly 1,500 full-time-equivalent decline in quarterly internal staffing to the Credit Suisse integration. The reported material does not link the reduction to AI.
What remains unknown about UBS’s AI deployment?
The bank has not disclosed a detailed definition of an agent, its data access and action permissions, how many employees use the tools, or how agent outputs are reviewed. It also has not quantified the productivity gains cited for developers.
Source: ThorstenMeyerAI.com
