🔍 Read the full analysis: The Questions UBS’s AI Agent Rollout Leaves Open on ThorstenMeyerAI.com
Get smart everyday buys delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
UBS reported in its Oct. 28 third-quarter results that employees run more than 12,700 AI personal agents, alongside 608 AI applications in production and more than 1,090 in development. The bank has not disclosed what the agents can access, whether they can take actions without human review, or how they are monitored.
UBS said employees are running more than 12,700 AI personal agents to support their daily work, reporting the figure in its third-quarter results on Oct. 28. The bank also listed 608 AI applications in production and more than 1,090 in development, but did not explain what its personal agents can access, what actions they can take or how the bank oversees them.
The figures appeared in a paragraph about investment in UBS’s business, within a 16-page results release otherwise focused largely on financial performance, capital requirements and the Credit Suisse integration. UBS said it has nine large-scale AI initiatives, and that live AI use cases were up 81% from a year earlier. It also reported more than 18,000 software developers using AI tools, saying those tools had led to measurable productivity gains without quantifying the gains.
The bank said its STAAT tool had generated more than 30 million insights so far this year for US financial adviser teams. Reuters and PYMNTS reported the 12,700-agent figure after UBS’s results, but did not provide further operational detail. UBS’s release does not define “personal agent” or say how many employees use them.
UBS’s second-quarter results on July 29 said all employees had the ability to create their own AI agents. In the three months since, the bank’s reported live use cases rose from 560 to 608, while applications in development increased from more than 920 to more than 1,090. Those figures describe separate categories of AI activity; UBS did not say that every use case or application is an agent.
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
Why Agent Oversight Matters at UBS
The disclosure offers a concrete measure of AI deployment at a large, regulated bank, but the count alone does not show how much work the systems perform or what risks they pose. A tool that drafts a summary for an employee differs from one that can send a message, approve a request or change a client account. Capabilities and permissions matter as much as the number of agents.
That distinction is relevant beyond UBS. Companies adopting employee-built agents face questions about access to internal information, human review, accountability for mistakes and the ability to track tools created across teams. UBS’s figures make the scale visible; without corresponding information about controls and outcomes, they do not establish how safely or effectively the agents operate.
The bank’s release also presents adoption alongside productivity claims, but gives no numerical estimate of the gains attributed to its more than 18,000 developers using AI tools. Readers therefore cannot compare the reported scale with a quantified result such as time saved, costs reduced or improved service. The deployment count is not, by itself, evidence of a specific business impact.
Products Worth Considering
FLEXISPOT Standing Desk with Monitor Riser & Cable Management
Dell P2725H 27-Inch Full HD Monitor
Dell 24 Plus Monitor S2425HSM – 23.8 FHD (1920×1080) 144Hz 1ms, 2 x 3W Speakers, HDMI, Height/Tilt/Pivot/Swivel Adjustability, AMD FreeSync
As an affiliate, we earn on qualifying purchases.
UBS’s AI Rollout and Client Boundary
In July, UBS said all employees could create their own AI agents, describing the effort as a way to deploy AI across the firm while freeing employees for judgment, collaboration and client work. UBS’s technology materials also describe an in-house assistant called Red, rolled out to around 100,000 employees, and an HR assistant that can help staff book holidays and find policies in Microsoft Teams. These examples provide some background on workplace AI at the bank, but UBS has not said whether either system is included in the 12,700-agent count.
UBS has drawn a separate boundary around external models acting for clients. Reuters reported that CEO Sergio Ermotti said at an ETH Zurich event on Oct. 27 that the bank was not allowing outside AI models to enter its systems and act on clients’ behalf, such as managing accounts or moving deposits. That statement addresses a particular use of external models; it does not describe the permissions or controls for employee-run agents inside UBS.
UBS also reported that its internal staff fell by about 1,500 full-time equivalents during the quarter, to 97,593. UBS and Reuters linked the reduction to the Credit Suisse integration. UBS has not attributed the staff change to AI, and the reported figures do not establish a connection between workforce changes and agent use.
“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 Disclosed
UBS has not explained what it means by a “personal agent,” which models or platforms power the tools, or what internal systems and data they can access. The bank also has not said whether agents can carry out actions independently or only prepare material for an employee to review. Those details determine whether an agent is functioning mainly as a drafting assistant or can initiate consequential activity.
The results do not specify how agents are approved, tested, monitored or shut down; how frequently employees use them; or how many of the 608 production AI use cases involve clients. The bank has also not reported error rates, incident figures or a quantified measure of productivity attributable to the agents. The source material does not establish how the 12,700 figure was counted or whether it refers to tools created, currently active agents or another measure.
Swiss regulator FINMA expects banks to maintain a complete inventory of AI applications and assign risk classes. It has said inventories have become harder to maintain as applications become accessible to more people. That regulatory expectation does not, on its own, show whether UBS’s inventory is complete or how it classifies these agents. No further information in the cited reporting resolves those questions.
Details Needed on Controls and Use
The next useful disclosures from UBS would explain the definition behind its agent count, the tools and data agents can reach, whether they can take actions and what review is required before those actions occur. The bank could also clarify how it classifies and monitors employee-created agents, and provide measured outcomes behind its statements about productivity gains. Those details have not been provided in the reported results.
Investors and employees may look for additional information in UBS’s future results, technology disclosures or comments from its executives. The current material does not identify a specific date for a fuller account of the agents or promise that one will be published. Until then, the reported scale is clear, while the systems’ precise role, oversight and impact remain open questions.
Key Questions
What did UBS report about its AI agents?
In its Oct. 28, 2026, third-quarter results, UBS said employees were running more than 12,700 AI personal agents to support their daily activities. The bank did not define the term in the release.
What can the UBS agents do?
The results do not specify their capabilities. UBS has not said what data or systems the agents can access, or whether they can take actions independently rather than prepare work for an employee.
Can outside AI models act on UBS clients’ accounts?
Reuters reported that CEO Sergio Ermotti said on Oct. 27 that UBS was not allowing outside AI models to act for clients, including managing accounts or moving deposits. That statement concerns outside models and does not detail the capabilities of UBS employees’ personal agents.
Did UBS link its staff reduction to AI?
No. UBS’s internal staff count fell by about 1,500 full-time equivalents during the quarter, to 97,593, and UBS and Reuters attributed the reduction to the Credit Suisse integration. UBS did not link the change to AI in the cited material.
Source: ThorstenMeyerAI.com
