Inside look: Citi’s ‘Four S’ AI strategy to maximize ROI
NEW YORK — Citi is deploying AI not as an additional layer in its tech stack but rather building its modern applications on top of a centralized platform.
New tech deployment must generate value at Citi, Head of AI Shobhit Varshney, said at Reuters’ Momentum AI Finance USA conference on Nov. 17.
“The definition of value comes from growth,” he said.
Varshney explained that the $1.7 trillion bank has a “Four S” strategy to deploy AI to generate maximum return on investment:
- Scale (organizational size and data volume): Citi moves more than $5 trillion a day and employs more than 230,000 people in more than 80 geographies, Varshney said, adding that any tech that is deployed needs to be provided to all employees.
- Speed (rapid deployment and adoption): With global operations, Citi prioritizes deploying AI quickly, he said. If the bank can roll out tech quickly, operations are more agile.
- Skills (talent development and training): The bank is continuously investing in its team members to get the most out of AI, he said. “We have done a ton of work in making sure that we see AI as an amplifier for human intelligence,” Varshney said. “A lot of the training has gone into the tools that have been rolled out over time.” The bank has developed Citi Assist, an internal gen AI-driven knowledge finding tool that can aid employees with tasks they perform, he said, adding that employees were taught how to use the tool and what functions Citi Assist can perform.
- Safeguards (governance and risk management): The bank is keeping safety at the forefront of AI development and deployment, Varshney said. This allows Citi to take on risky projects with ease because it has a framework for using AI while avoiding hallucinations common with gen AI.
As part of a multiyear modernization, the bank has been preparing for broader deployment of AI, he said, adding that it is already seeing greater efficiency in specific areas.
“We have about 30,000 developers, and we have unlocked about 100,000 hours of [additional] productivity every week,” Varshney said. “Now we can do over a million code reviews fully automated every week.”
The bank is seeing high productivity gains from AI in wealth management and customer service, Varshney said, adding that gen AI can create reports and complete documentation.
“For every line of business, every functional area, we have figured out where we are spending most of our effort … and are directing AI deployment to those areas,” he said.
The quality of AI output has improved significantly, Varshney said. In its next phase of AI deployment, Citi will look toward AI to generate revenue.
Register here for early-bird pricing for the inaugural FinAi Banking Summit 2026, taking place March 2-3 in Denver. View the full event agenda here.
