We are experiencing some temporary issues. The market data on this page is currently delayed. Please bear with us as we address this and restore your personalized lists.
Unlock stock picks and a broker-level newsfeed that powers Wall Street.
How Citi’s CTO is rolling out new gen AI productivity tools to more employees across the globe
Rohit Bhat, managing director for financial services at Google Cloud (left) and Citigroup Chief Technology Officer David Griffiths. ·Fortune·Photos courtesy of Google, Citigroup
At Citigroup an "AI accelerator" is not a piece of hardware, it's a person. Citi's accelerators—a diverse cohort of employees from all parts of the company—are familiar and enthusiastic about using AI tools, and play a key role in helping the bank adopt the technology across the organization.
On Wednesday, Citi gave a glimpse of this process in action as Chief Technology Officer David Griffiths announced that the company was expanding the use of several generative AI tools for its workforce to a total of 11 countries, up from eight countries at the end of 2024.
The AI tools, which include a virtual assistant and a coding assistant, are getting some important improvements that Griffiths says were derived from the accelerators' feedback and from analyzing real world data from all the workers using the technology in the initial regions.
“That's very much been our strategy: Get the tools out at scale so that as they advance, you maximize the benefit and the impact,” says Griffiths. Thus far, he says Citi has focused on markets that have the most stable regulatory environments, including the U.S., U.K., India, Canada, Costa Rica, and Poland.
Citi has a physical presence in close to 90 countries globally and while Griffiths says that he can’t be certain generative AI will be unveiled in all regions, he expects a “large majority” of the bank’s 230,000 global workforce will have access to these tools in the coming months.
Last year, the bank launched Citi Assist, a virtual assistant that can help answer policy and procedural questions, and an AI tool called Citi Stylus, which summarizes documents, performs translation, and can compare multiple documents at once. The new version of Citi Stylus, which rolled out to around 10,000 early adopter Citi employees last week, allows them to ask more detailed follow up questions and edit and store the outputs from the large language models.
“It’s broader, it’s more interactive, and the outputs that it generates are more immediately consumable than before,” says Griffiths, who shared the fresh product updates at a session held Wednesday at a Google Cloud conference held in Las Vegas.
Citi also debuted an AI coding tool for developers, called Citi Squad, in 2024 and made it available to about 9,000 employees. Assist and Stylus have been rolled out to 150,000.
The improvements made to Stylus reflect Citi’s data-driven approach of actively tracking how employees were using the tool, as well as employee feedback that can be shared directly within the tool and, of course, the insights provided by Citi's AI accelerators.
Citi’s generative AI efforts lean on a close partnership with Google, including a multiyear partnership announced last year to move parts of the bank’s financial infrastructure to Google Cloud. Through that partnership, Citi has access to Vertex AI, a singular AI platform used to build AI applications like Assist and Squad and allowing the bank to train and manage LLMs, including Google’s proprietary Gemini.
“We can all hypothesize when new technologies come out where we think value will be, but the fastest way to ensure that it's credible is to get a feedback-driven system in place as quickly as possible,” says Rohit Bhat, managing director for financial services at Google Cloud. “Not just with the operators, not just with the developers, but with the end users.”
Citi's Griffiths says there’s so much innovation in the market that he wanted the optionality of a multi-modal strategy from a variety of AI hyperscalers, though his strong preference is for closed-source, which restricts access to the source code.
Beyond bolstering employee excitement about generative AI, Griffiths is also working to integrate the tools more seamlessly in workflows. Citi Stylus and Citi Assist were both built as standalone, separate applications, but that’s changing too. There is now a web plugin for Citi Stylus so that employees can summarize and create Q&A’s when browsing the internet, as long as the page isn’t paywalled. Citi Assist is being added to the various chat apps that Citi employees are already using from vendors.
Griffiths is tracking AI progress by how much “capacity” Citi is creating, which he explains is a usage-based metric. When Citi launches a tool, it looks at if a human did the work 100 times, and it costs “X,” and for AI it costs “Y,” Citi can then calculate the potential financial gains that can be accrued by launching these productivity tools.
Griffiths is also developing multi-step agentic workflows that could, in one example, take a summary, translate it, and create presentation slides. The technology isn’t reliable enough yet for prime time, but he says the large language models’ reasoning capabilities are getting better. “The agentic model is becoming much more viable,” says Griffiths, who expects the most significant gains will be with developers.
Citi, which spends about $12 billion annually on technology, says its generative AI rollout has prioritized tools that can be launched horizontally across all parts of the business, including wealth management and front-office banking. But a parallel path has emerged where Citi will also develop job-specific AI tools that can be tailored differently for developers, customer service, sales and marketing, and more.
“We need a set of targeted solutions, but ultimately the end result is the same,” says Griffiths. “There are things that people are doing today that they will be much more efficient at doing in the future.”
John Kell
Send thoughts or suggestions to CIO Intelligence here.
