Citigroup AI Strategy: AI at Scale on a Rebuilt Foundation
How Citigroup approaches AI: AI tools for 180,000 employees, required prompt training, one Group Head of AI, an agent platform and 100 processes rebuilt. Updated November 2026.
Global bank: services, markets, banking, wealth and US personal banking
Citi sets out its AI plans within its corporate strategy, in its 2025 annual report and earnings calls. It has no standalone AI strategy document.
Read the company sourceWhere Citigroup Is Going with AI
Citi is putting AI to work across the whole bank, on top of the systems and data it rebuilt over five years of transformation. More than 180,000 employees have its AI tools and more than 87% use them. A single Group Head of AI now runs one AI mandate across technology and operations, and Citi says about 100 of its most manual processes are being rebuilt with AI and automation. It reports capacity and speed gains, and says AI benefits outweigh the cost this year, but gives no dollar figure.
AI follows the transformation. CEO Jane Fraser told investors in October 2026: “Our transformation has modernized our core technology platform, giving us the foundation we need to implement AI firmwide” (third-quarter 2026 earnings call). That transformation began after regulators ordered Citi in 2020 to fix its risk management and data. Chief operating officer Anand Selvakesari puts the link plainly: “When data is trusted and governed, AI becomes significantly more effective” (Citi, October 2026).
The ambition is to lead. Group Head of AI David Griffiths says: “The banks that get this right first will be those that define the next era of financial services” (Citi, September 2026). Fraser says Citi is “deploying AI at scale with measurable results” (October 2026).
The focus has four parts, which Fraser calls growth, efficiency, employee training and defense (October 2026):
- Growth: faster product launches and Citi Sky, an AI adviser now in a pilot in Wealth (earnings call), plus AI insights for wealth advisers (2025 annual report)
- Efficiency: about 100 of the most manual processes rebuilt end to end with AI and automation, and AI coding tools for developers (earnings call; 2025 annual report)
- People: AI tools and required prompt training for the workforce (Fortune, January 2026)
- Defense: hardening systems and stress-testing cyber defenses (earnings call)
What It Takes: One AI Owner, Required Training and an Agent Platform
One owner, inside operations. In May 2026, Citi made chief technology officer David Griffiths its Group Head of AI. He sits in the chief operating officer’s leadership team rather than in technology, and works with Tim Ryan, Head of Technology and Business Enablement. Selvakesari said the move brings “clearer focus and acceleration” to Citi’s AI strategy, and Griffiths said: “Our new organisation unites Technology and Operations under a single, unified AI mandate” (AIM, May 2026). Citi hired Brian Saluzzo from Google as chief information officer in March 2026 (CIO Dive), and in October a Google engineering director to lead its agentic AI frameworks (American Banker, October 2026).
One platform, then agents.
- Tools for everyone: nearly 180,000 colleagues in 87 jurisdictions have Citi Stylus Workspaces, Citi Assist and Microsoft 365 Copilot Chat (Citi, October 2026). Stylus Workspaces added agentic features in September 2025 (Citi release).
- Agents: Arc, launched in April 2026, lets Citi build AI agents and embed them in everyday work. Griffiths said: “For the first time, we can deploy embedded AI agents at enterprise scale across every business line” (CIO Dive, May 2026). It is the step across the chat-to-agent gap.
- Coding: 14,000 engineers use Devin, an AI coding agent (October 2026), which Citi says mapped more than 30 years of legacy code in two days (Citi).
Training is required. About 175,000 employees in 80 locations had to complete AI training focused on prompting, which adapts to each person: under 10 minutes for experts and about 30 for beginners (Fortune, January 2026). Most employees have completed it (CIO Dive, May 2026). Fraser says: “I want to stack the odds that we will help our people reinvent themselves” (Fortune). Prompting is the entry point; it is becoming a core skill, not a specialist job.
Adoption spreads through people. Citi built a network of more than 4,000 AI Accelerators inside teams, supported by 25 to 30 AI Champions. We covered how Citi scaled AI adoption socially, and Citi is one of our AI implementation case studies.
Controls are built into the platform. Griffiths says “controls are automated at the platform level,” across the whole AI lifecycle, with a human always in the loop (Citi, September 2026). Citi is also adding continuous monitoring and testing of its controls (Citi, October 2026).
The workforce is shrinking. Citi had about 226,000 employees at the end of 2025 and about 219,000 at the end of June 2026 (second-quarter call). Its restructuring, announced in 2024, could remove up to 20,000 roles (Fortune, January 2026). Fraser has said that some roles will change, new ones will appear and “others will no longer be required” (Fortune). In July, CFO Gonzalo Luchetti said Citi may take more severance in the second half as AI and automation speed up productivity gains (second-quarter call), and by September it had booked more than $1 billion of severance in 2026 (third-quarter call).
AI fluency, coded against the Lead with AI Fluency Matrix (September 2026 edition):
- Fluency floor: Level 1 (Chatting), by requirement. Prompt training is required, and Citi reports that more than 87% of employees with its AI tools use them.
- Highest level evidenced: Level 3 (Integrating), company-claimed. Agents are built on Arc and embedded in workflows, 14,000 engineers use a coding agent, and AI supports client service in 82 countries. Citi says about 100 processes are being rebuilt end to end, but the sources reviewed show no single process run by AI end to end with a measured result, so Level 4 is not evidenced.
- Portfolio design: Platform-led and firmwide. One set of AI tools for all staff, one agent platform, a central AI team in the chief operating officer’s organization, a network of AI Accelerators in teams, and controls built into the platform.
How Citigroup Measures Results
Citi reports adoption, capacity and speed. All figures below are company-claimed:
- Adoption: more than 87% of employees with AI tools use them, with more than 65 million interactions since the rollout began in December 2024 (Citi, September 2026)
- Developers: AI coding tools for 30,000 developers create about 100,000 hours of capacity a week (2025 annual report)
- Speed: product launches 30% to 40% faster, and client response times in Services 21% better (October 2026)
- Processes: in US client onboarding, document review fell from about an hour to 15 minutes (Citi), and about 100 manual processes are being rebuilt end to end (October 2026)
The payoff has no number yet. Fraser says: “This year alone, we’re seeing the financial benefits from AI and automation significantly outweigh the investments” (October 2026), but Citi gives no figure for either side. CFO Luchetti says the chief operating officer and the head of technology review the 100-plus processes every week (July 2026). Putting a value on them is the step from AI adoption to AI value.
Next markers: a dollar figure for AI benefits, how many of the 100 processes run end to end on AI, how far agents on Arc spread beyond developers, and whether Citi Sky leaves its pilot.
Change log
- December 2024: Citi begins rolling out its internal AI tools to employees.
- September 2025: Citi adds agentic features to Stylus Workspaces.
- October 2025: Citi requires prompt training for about 175,000 employees.
- April 2026: Citi launches Arc, its platform for AI agents.
- May 2026: David Griffiths becomes Group Head of AI.
- October 2026: Fraser says AI benefits outweigh the investment this year; Citi has booked more than $1 billion of severance in 2026.
The Bottom Line
- Disclosure: 4 of 4. Citi answers where it is going, where it will focus, what it will take and how it will know, with adoption, capacity and speed figures. It gives no dollar value for AI.
- AI rides on the cleanup. Citi built its AI push on the data and platforms it rebuilt during its transformation.
- Adoption is near-universal, with required prompt training, more than 4,000 AI Accelerators in teams and more than 87% use.
- One owner, in operations. The Group Head of AI sits with the chief operating officer, not in technology.
- The say-do gap is the workforce. Fraser speaks of helping people reinvent themselves, while the CFO ties more severance to AI and automation and headcount keeps falling.
- Watch the number. The test is whether Citi puts a dollar figure on the benefits it says already exceed the cost.
Revenue, Headcount and Revenue per Employee
What the numbers show: Revenue rose 14% from 2022 to 2025 while headcount fell 6%, so revenue per employee rose 21%, to about $377k. Headcount kept falling in 2026, to about 219,000 by June, and Citi booked more than $1 billion of severance in the first nine months. These figures do not show that AI caused any change.
Revenue, headcount and revenue per employee, FY2022 to FY2025
- FY2022: $74.98bn revenue, 240,000 employees, $312k per employee
- FY2023: $78.07bn revenue, 239,000 employees, $327k per employee
- FY2024: $80.72bn revenue, 229,000 employees, $352k per employee
- FY2025: $85.23bn revenue, 226,000 employees, $377k per employee
Change, FY2022 to FY2025: revenue up 14%, headcount down 6%, revenue per employee up 21%.
Notes: Revenue is revenues, net of interest expense, as restated in Citi’s 2025 annual report. From 2025, Citi nets certain fees paid to card networks against revenue and applied this to all four years, which lowers revenue by about $0.4 billion a year against first reports. Headcount is direct staff at year-end, rounded to the nearest thousand. Citi had about 219,000 employees at the end of June 2026.
These figures give context on scale and productivity. They do not show that AI caused any change. Sources: Citi 2025 annual report and 10-K filings.
How Citigroup Compares with Financial Services Peers
We looked at how Citigroup compares with other financial services companies in the index. Here is how they stack up.
| Company | Disclosure | AI owner | Fluency floor | Governance | AI value reported | Revenue per employee |
|---|---|---|---|---|---|---|
| Allianz | 3 of 4 | No named AI owner; board COO covers IT | Level 1, by participation | Eight principles and a register of every use case | Results by use case, no total | $1.35M |
| American Express | 3 of 4 | EVP, Enterprise Data and AI | Not published; tools for nearly all colleagues | Agents only through approved API or MCP channels | No: coding time only | $940k |
| Bank of America | 4 of 4 | CTIO; no Chief AI Officer | Not disclosed | Controls, no published principles | Yes: $250M to $300M in 2026, with a 2027 figure | $531k |
| BBVA | 4 of 4 | AI Transformation area at the top of the bank | Level 1, training required before access | Eight published principles | Results by process and agent | $328k |
| Berkshire Hathaway | 1 of 4 | None at group level; each business decides | Not published | Not published | None published | $958k |
| BlackRock | 2 of 4 | No firmwide AI leader named; Aladdin engineering builds the tools | Not published | Guardrails built into the agent platform | None published | $973k |
| BNY | 4 of 4 | Chief Data and AI Officer | Level 1, near-universal training | Every digital employee has a human manager; reviews before release | Speed results by use case, no total | $417k |
| Capital One | 3 of 4 | Chief Scientist and Chief Enterprise AI Officer | Not published | Every agent evaluated continuously, with human oversight | No: declines to quantify benefits | $700k |
| Charles Schwab | 3 of 4 | CIO for technology, operations and data; no AI head named | Not published | Guardrails and human handoffs; outside agents cannot move money | No: counts of use cases only | $725k |
| Citigroup | 4 of 4 | Group Head of AI, in the COO’s organization | Level 1, required prompt training | Controls built into the AI platform, human in the loop | Capacity and speed results; says benefits exceed costs, no figure | $377k |
| Goldman Sachs | 3 of 4 | CIO, sponsored by the top three executives | None declared | Firmwide AI risk committee | No: weekly use only | $1.23M |
| JPMorgan Chase | 4 of 4 | Chief Data and Analytics Officer, passing to the CTO | None declared (use is opt-in) | Little published | Earlier savings estimate, no longer updated | $573k |
| Lloyds Banking Group | 4 of 4 | Chief Data and AI Officer | Level 1, mandatory training | Responsible AI team and framework | Yes: about £50M in 2025, £100M+ target | $377k |
| Mastercard | 3 of 4 | Chief AI and Data Officer | Level 1, data training required before access | Every model explainable, governed and monitored | No: few internal results | $824k |
| Morgan Stanley | 3 of 4 | Head of Firmwide AI | Not published; 98% of adviser teams use the AI assistant | Evaluations before every launch; human oversight | Results by use case, no total | $851k |
| U.S. Bancorp | 4 of 4 | Dedicated AI teams; no single AI leader named | Not published; AI courses offered to all | Not published | Results by use case, no total | $417k |
| Visa | 4 of 4 | President of Technology | Not published; generative AI access for all since 2023 | Data-use principles built into model monitoring | Product development results, no total | $1.32M |
| Wells Fargo | 3 of 4 | Head of AI, reporting to the CEO | Not published; 90,000+ trained | Business heads accountable; rules for customers’ AI agents | No: coding efficiency only | $408k |
| Klarna AI Pacesetter | 3 of 4 | CEO | Not disclosed | None published | Yes: savings and work done by AI | $1.24M |
Disclosure counts how many of the four strategy questions (where it is going, where it will focus, what it will take and how it will know it is working) a company answers in public. Fluency floor is the lowest level of AI fluency a company expects of all its people, on the five levels of the Lead with AI Fluency Matrix: Avoiding (0), Chatting (1), Collaborating (2), Integrating (3) and Transforming (4). The other columns summarize each company’s profile. Revenue per employee is for each company’s latest fiscal year, in US dollars. Rows marked AI Pacesetter or Notable are companies outside the S&P 100 and STOXX Europe 50. They are shown for contrast and are not counted in any index figure. See all companies in the AI Strategy Index.
Frequently Asked Questions
What is Citigroup’s AI strategy?
Citi is putting AI to work across the whole bank, on the systems and data it rebuilt during its transformation. CEO Jane Fraser describes four aims: growth, efficiency, employee training and defense.
Who leads AI at Citi?
David Griffiths, Group Head of AI since May 2026 and previously chief technology officer. He sits in the leadership team of chief operating officer Anand Selvakesari and leads one AI mandate across technology and operations.
How is Citi training employees in AI?
Citi required about 175,000 employees to complete adaptive AI training focused on prompting, starting in 2025. More than 87% of employees with Citi’s AI tools use them.
Is Citi cutting jobs because of AI?
Citi’s restructuring could remove up to 20,000 roles, and headcount fell to about 219,000 by June 2026. Citi has not tied a number of job cuts to AI, but its CFO said in July 2026 that more severance may come as AI and automation speed up productivity gains.
What results does Citi report from AI?
Citi reports about 100,000 hours of developer capacity a week, product launches 30% to 40% faster and client response times 21% better in Services. Fraser says AI benefits outweigh the investment in 2026, but Citi has published no dollar figure.
How is this profile put together?
Lead with AI profiles use public sources only, such as annual reports, earnings materials, filings and on-record interviews, dated at the time of writing. AI fluency is coded against the Lead with AI Fluency Matrix. Company figures are marked as company-claimed unless they have been independently verified.
About the AI Strategy Index
The Lead with AI AI Strategy Index profiles how the largest companies in the US and Europe approach AI, starting with the S&P 100 and the STOXX Europe 50. Each profile answers four questions: where the company is going with AI, where it will focus, what it will take and how it will know it is working. AI Pacesetters and notable companies outside the two indexes are profiled for contrast and are not counted. Profiles use public sources only, and AI fluency is coded against the Lead with AI Fluency Matrix.
See all company profiles