"The platform, known as Citi Stylus Workspaces, uses a range of models, including Google's Gemini and Anthropic's Claude. One example Griffiths cited: a user can instruct the system to research a client, compile a profile using both public and internal data, and translate the results into another language - all in a single step. Previously, each of these actions required separate user input."
"The initial pilot will involve 5,000 users over a period of four to six weeks. Griffiths said the goal is to measure usage patterns, impact on productivity, and the cost-to-value ratio. AI agents can become expensive quickly as they consume tokens - units of model usage - especially for longer, more complex tasks. To address this, Citi has implemented strict cost controls. Griffiths noted that rapidly falling model prices make ROI calculations challenging, but most tasks in this pilot should complete in minutes, keeping costs manageable."
Citi Stylus Workspaces integrates multiple large models, including Google's Gemini and Anthropic's Claude, to enable single-prompt execution of multi-step corporate tasks. The system can research clients, compile profiles using public and internal data, and translate outputs in one automated sequence instead of requiring separate commands. An initial pilot will involve 5,000 users over four to six weeks to measure usage patterns, productivity impact, and cost-to-value. Strict cost controls are in place to limit token consumption and keep tasks, which typically finish in minutes, economically manageable amid rapidly falling model prices. The capability has potential to reshape how routine corporate work is executed.
Read at Medium
Unable to calculate read time
Collection
[
|
...
]