UBS Introduces AI Clones of Its Analysts
Zurich-based Swiss global bank UBS has begun deploying AI avatars of its analysts to meet growing client demand for research content delivered in video format. This pioneering initiative uses advanced models from OpenAI and Synthesia to create realistic digital versions of UBS experts, representing a significant step forward in the use of artificial intelligence in financial services.
So How Does It Work?
The process of creating these AI avatars is carefully designed to preserve the authenticity and quality of the research content. Analysts participate voluntarily in a process in which they record their likeness and voice in a studio using Synthesia technology. The system then uses language models, such as those from OpenAI, to analyze the analyst's written reports and generate a script for the video. The entire process is based on transparency and quality control, with analysts required to approve both the script itself and the final video before distribution. All content featuring an AI avatar is clearly labeled as AI-generated throughout the video. Scott Solomon, head of global research technology in UBS's investment banking division, identified two main factors that led to the adoption of this technology. The first is growing client demand for consuming research content through short videos—a trend influenced by platforms such as TikTok. The second factor is the operational efficiency enabled by the avatars. They allow analysts to focus more on their core research work while still meeting clients' needs for timely video updates.
More Videos and, of Course, Security
The significant improvement in efficiency is evident from the figures released by UBS. Traditionally, the bank produced approximately 1,000 research videos per year due to studio limitations, despite publishing around 50,000 documents annually. With this new technology, UBS aims to scale production to approximately 5,000 avatar videos per year, representing a fivefold increase in production capacity. The bank places considerable emphasis on transparency and security measures. As Scott Solomon emphasizes: "Throughout the video, you will see a notice at the bottom stating, 'this content was created using AI.' We are in no way trying to replace a flesh-and-blood analyst." Participation in the program is strictly voluntary for analysts, and all use cases require explicit approval from both human resources and the individual analysts before deployment or publication.
However, the implementation has not been without technical challenges. The rollout encountered some issues, particularly with accurately replicating certain accents, which sometimes resulted in less authentic-sounding avatars. In response, UBS slowed the rollout for affected employees until these issues were resolved, demonstrating the bank's commitment to maintaining the high quality and authenticity of its output. This move by UBS reflects a trend among major financial institutions experimenting with generative AI tools to reduce costs and increase efficiency while proactively responding to changing client media consumption habits. While concerns about deepfakes persist elsewhere, particularly regarding the misuse of celebrities' likenesses, UBS maintains strict controls over authenticity disclosures and does not attempt to pass the avatars off as real people.
