In 2020, U.S. regulators fined Citibank (Citigroup’s subsidiary) $400 million for deficiencies in enterprise-wide risk management, data governance, and internal controls.
The same year, the bank accidentally wired $900 million to Revlon’s lenders due to a system error. In 2024, another glitch caused an $81 trillion crediting error.
CEO Jane Fraser called the root cause decades of underinvestment in tech and infrastructure.
Despite being one of the largest financial institutions with roughly 200 million customer accounts, Citigroup had to rebuild itself to get on track with the AI world.
In today’s AI at the Top, we will learn how Citigroup went from regulatory issues and poor infrastructure to an over 70% AI adoption rate within the company.
Why Citigroup Had to Fix the Foundation Before Scaling AI
The company has businesses across consumer banking, wealth management, institutional trading, treasury services, and investment banking.
The bank’s systems had been built over decades, with different teams solving the same problems in different ways. Data and applications were not interconnected. Regulators flagged data quality as a critical weakness.
It was important for different domains to learn from each other, so the models can find patterns and optimise for efficiency.
CEO Fraser’s response to poor infrastructure was a three-part strategy.
Simplify the business. Citigroup exited consumer banking in 14 international markets to reduce organisational complexity. It made the bank focus on high-margin businesses like institutional banking and wealth management. AI delivers most value here because of the data richness.
Retire legacy tech. Since 2022, the bank has removed more than 2,000 legacy applications. This turned away the systems that caused past failures and created a cleaner base for AI integration.
Build a modern data and cloud layer. Citigroup partnered with Google Cloud, gaining access to Vertex AI to build its own tools without developing foundational models from scratch. The bank implemented a global data governance framework, making sure information from trusted sources can’t be altered.
Andy Sieg, Citigroup's wealth chief, said the rebuilt architecture lets the bank make progress in months, which once would have been years.
Enjoying this read? Share the smartest part with your followers. Click to Share.
How Citigroup Built an AI-First Culture Across 180,000 Employees
Citigroup knows it’s not the tools, but talent.
So it has built 4000 AI champions across the company, who are enthusiastic about AI and want to help their peers adapt to tech.
These aren’t formal trainers but colleagues who use the same tools in their everyday workflows.
Leadership’s message to employees was “just jump in and start exploring.” In September 2025, the bank rolled out mandatory prompt training to 175,000 employees in 80 locations.
Employees have pitched over 350 generative AI use cases so far and the adoption has crossed 70%.
With the culture in place, Citigroup deployed four core AI tools across the enterprise:
YOU DECIDE
🌟 Want to be featured in the next issue? Reach out with your best AI use case and we’ll spotlight it.
🏢 What company do you want us to cover next?
Citigroup's AI Impact on Its Business
With the right infrastructure, data foundation, and adoption, the results are showing up in hard metrics across Citigroup's business lines.
In the Q4 2025 earnings call, CEO Fraser said, “AI is now driving efficiencies across multiple processes and creating new sources of efficiency that three, four years ago, we couldn’t have imagined.”
What's Next: Agentic AI and the “Do It For Me” Economy
Citigroup is already building for what comes after chatbots and copilots.
CTO David Griffiths mentioned the bank is moving towards autonomous, agentic use cases. Citi Stylus Workspaces, launched in September 2025, is the first step, an agentic platform where employees complete multi-step tasks with a single prompt.
The bank published a detailed report on agentic AI: ‘Do It For Me’ economy.
For banking, that means real-time portfolio rebalancing, automated invoice processing, and dynamic lending offers based on live cash flow data.
We are all moving from AI helps you work to AI works for you.
What Enterprise Leaders Can Learn from Citigroup's AI Strategy
Fix the foundation before scaling AI. Citigroup spent $30 billion rebuilding infrastructure and retired 2,000 legacy apps before deploying AI at scale. The tools work because the data layer underneath them is clean and unified.
Treat adoption as a culture problem, not a tech rollout. The 4,000-person AI Champions network, mandatory prompt training, and a clear encouragement from leadership drove 70%+ adoption. Tools don't create change. People do.
Measure capacity created, not logins. CTO David Griffiths tracks the delta between human-only effort and AI-assisted effort across sample tasks. That keeps the focus on economic value, not vanity usage metrics.
Sources:
YOUR TURN