Citi Rebuilds Its AI Leadership for the Third Time in a Year
Citigroup has rebuilt the leadership of its artificial intelligence organization for at least the third time in roughly thirteen months. This week it named Sebastian Gnagnarella, a former Google director of engineering who oversaw enterprise deployments of Gemini, to a newly created post: head of applied AI and what the bank is calling its 'agent factory.' Three existing executives were reassigned around him in the same memo, according to American Banker's reporting, corroborated by South Korean outlet Digital Today.
A Chair That Keeps Changing Hands
The churn predates this week. Citi hired Shobhit Varshney away from IBM Consulting in September 2025 to be its first global head of AI, a move CIO Dive and Banking Dive covered as a marquee signing meant to centralize the bank's AI push. Varshney lasted six months before leaving, citing personal reasons, as Risk.net reported in April 2026. Citi didn't name an immediate successor; instead, in May, it elevated CTO David Griffiths into a new 'Group Head of AI' title sitting inside the chief operating office rather than the tech org, a structural choice that outlasted exactly one round of this week's changes before being supplemented again.
What's New This Time
Gnagnarella's 'agent factory' title suggests Citi wants a dedicated owner for building and deploying AI agents across the bank, distinct from the governance and strategy work being split off to others: Aida Filipovic now leads responsible AI (safety, ethics, governance), Niamh O'Sullivan takes AI strategy, and Sheetal Chanderkar becomes chief operating officer for the AI organization itself — a COO for the AI unit, layered under Griffiths, who is himself nominally the bank's top AI executive. It is also Citi's second AI-adjacent hire from Google in seven months, after CIO Brian Saluzzo joined from the same company in March.
The Unanswered Questions
Citi hasn't explained, on the record, why Varshney left after half a year, nor whether this week's additions represent net new headcount in AI leadership or simply another rearrangement of existing roles under a different org chart. The bank's claim that AI has sped up product launches by 30 to 40 percent, attributed to CEO Jane Fraser, is self-reported and comes with no methodology attached. What is independently verifiable is the hiring pattern: industry data cited alongside the Citi reporting puts AI-related job postings in banking up roughly 49 percent year over year, meaning Citi is one of many banks bidding for the same small pool of people who've actually shipped generative AI products at hyperscale.
Taken together, the pattern reads less like a single bold AI strategy and more like a bank still deciding, in public, where AI leadership should sit — under the COO, under a dedicated agent-building unit, or spread across specialists in strategy, ethics and operations. Three structures in thirteen months is not necessarily instability; it could just as easily be a bank iterating quickly on an unsolved org-design problem every large institution adopting agentic AI is now facing. Citi isn't saying which it is, and investors won't get much more clarity until the bank's next earnings call.