AI in Banking: Two Races Define the Financial Future According to Citi

What happened?

Jane Fraser, CEO of Citigroup, told the South China Morning Post that the financial sector is competing in two AI races at the same time. The first is offensive: applying AI to business models to accelerate product development, improve customer service, and increase revenue. The second is defensive: protecting the institution from the risks that AI itself can generate, such as advanced fraud, cyberattacks, or algorithmic biases. This statement, reported by The Next Web, is not an isolated comment; it reflects a strategy that Citi is already implementing. Fraser explained that the bank has created an AI governance committee and is investing in deepfake detection tools and explainable models. Additionally, Citi has allocated $1 billion to AI technologies by 2027, according to internal sources.

Why is it important?

Fraser's statement is not a mere opinion: Citigroup is one of the world's largest banks, with over 200 years of history and a presence in 160 countries. Its stance sets a trend. The offensive-defensive duality reflects a reality that many tech and financial companies face: AI is not only a growth tool but also a source of new vulnerabilities. According to a McKinsey report, generative AI could add between $200 billion and $340 billion annually to the global banking sector, but associated risks (regulatory, ethical, operational) could erode up to 20% of that value if not properly managed. For context, during the 2008 subprime mortgage crisis, banks lost over $2 trillion; now the risk is not just financial but systemic, as algorithms can amplify biases or facilitate fraud at scale. A study by the Bank for International Settlements (BIS) warns that AI could destabilize markets if its use in algorithmic trading is not controlled.

Consequences for the sector

Transformation of products and services

The offensive race means banks will integrate AI into processes such as credit origination, offer personalization, advanced chatbots, and spending pattern detection. JPMorgan Chase, for example, already uses language models to summarize legal documents, and Bank of America has Erica, its virtual assistant with over 10 million users. It is expected that within the next two years, 80% of banks will implement generative AI in at least one critical area. But not only giants: fintechs like Klarna already use AI to automate 85% of their customer service interactions, and Revolut has launched a GPT-based assistant for personal finance management. This is compressing product development timelines: what used to take months is now achieved in weeks. However, speed also brings risks: in 2024, an error in a European bank's generative AI model mistakenly approved high-risk credits, causing losses of €50 million.

Risk management and compliance

The defensive race addresses the need to monitor AI models to avoid biases, ensure explainability, and comply with regulations like the European Union's AI Act. Banks will need to invest in AI governance teams, algorithmic audits, and AI-powered fraud detection systems. According to Gartner, by 2026, 60% of large financial institutions will have an AI ethics committee. A concrete case: Banco Santander has developed an "explainable AI" system that allows regulators to understand why a credit was denied. Additionally, cyberattacks using generative AI are on the rise: according to CrowdStrike, deepfakes used in banking fraud grew by 300% in 2025. Citi, for its part, has implemented a real-time deepfake detection system for identity verification video calls.

Impact on employment

Automation of repetitive tasks (credit analysis, customer service, compliance) could reduce jobs but will also create new roles: prompt engineers, AI auditors, data specialists. An Accenture study estimates that AI could displace 12% of banking employees by 2030 but will simultaneously generate 8% new positions. However, the transition will not be easy: in 2025, German bank Commerzbank announced the layoff of 3,000 employees while hiring 500 AI engineers. The skills gap is critical: according to a LinkedIn report, demand for AI experts in banking grew 150% in 2025, while supply only increased 40%. This is driving up salaries: an AI engineer in banking can earn up to $200,000 annually in the United States.

Competitive implications

The offensive-defensive duality also redefines competition. Large institutions have an advantage in the defensive race because they can afford governance and compliance teams, but fintechs can be more agile offensively. For example, Brazilian fintech Nubank already uses AI to offer personalized credits in seconds, without physical branches. Meanwhile, traditional banks like HSBC are forming alliances with AI startups to accelerate their transformation. Regulation also plays a role: the EU AI Act will classify credit systems as high-risk, requiring external audits. This could make innovation more expensive in Europe compared to the US or Asia.

What readers should know

• AI in banking is not optional: it is a two-front race where falling behind in either can be fatal. As Fraser said, "you can't focus only on growing if you don't protect your home."
• Consumers will see faster and more personalized services, but they must also trust that banks protect their data and make fair decisions. Transparency will be key: according to a PwC survey, 73% of users distrust automated decisions if they are not explained.
• Regulators are watching closely: transparency and explainability will become the norm. The Fed, ECB, and PBoC have already issued guidelines on responsible AI in banking.
• Small fintechs can compete by adopting offensive AI, but defense requires scales that only large banks can afford. This could lead to sector consolidation, where big banks acquire fintechs for talent and technology.

"Artificial intelligence is not the future of banking; it is the present, and it is played on two simultaneous fronts," summarizes Jane Fraser. Citi's vision lays the groundwork for a decade of transformation where technology and trust must go hand in hand. But as the BIS warns, if banks fail to balance both races, we could face a new financial crisis driven by algorithms.