Citi exec: Generative AI is transformative in banking, but risky for customer support

Key Points:

  • The shift towards data-driven decision making at Citi due to generative AI
  • Categorization of generative AI projects at Citi into three main buckets
  • Citi’s cautious approach towards using generative AI externally for customer interactions


Generative AI has significantly impacted decision-making at Citi, with a notable shift towards data-driven strategies, as shared by Promiti Dutta, the head of analytics technology and innovation. Dutta emphasized the transformative power of gen AI, which propelled data and analytics to the forefront of the organization, leading to a surge in AI project proposals from employees. Citi’s technology leaders embraced this curiosity, focusing on meaningful outcomes and certainty in generative AI projects.


Noteworthy AI project priorities at Citi include enhancing agent support with large language models (LLMs) for call center operations, automating manual tasks like compliance document reviews, and implementing an internal search engine to facilitate data-driven insights among employees. Despite internal successes, Citi has chosen to be cautious about external applications of generative AI, such as customer-facing chatbots, citing potential risks and the need to maintain customer trust.


While some industries have integrated LLMs into customer interactions, Citi remains vigilant due to the importance of accurate customer communication. The bank currently utilizes pre-scripted NLP responses but acknowledges the potential for future external LLM usage, with a commitment to maintaining human oversight for safety and customer understanding. This approach contrasts with Wells Fargo, which has successfully implemented LLM-driven solutions like the Fargo virtual assistant for customer service.


Furthermore, generative AI has prompted Citi to reassess its cloud usage, exploring solutions like OpenAI’s GPT models on Azure, despite historical reliance on on-premise infrastructure. A dedicated internal task force oversees all generative AI projects, highlighting the bank’s emphasis on responsible implementation and customer safety. Microsoft’s Sarah Bird commends Citi’s cautious approach, underlining ongoing efforts to enhance AI precision and avoid errors in applications such as retrieval augmented generation (RAG).


The collaboration between Citi and Microsoft underscores a commitment to leveraging generative AI responsibly and effectively. As the industry continues to refine AI technologies, both companies are dedicated to mitigating errors and enhancing grounding accuracy in AI applications. The cautious yet proactive strategies adopted by Citi exemplify a best practice approach to navigating the evolving landscape of AI integration in banking operations.



Prompt Engineering Guides



©2024 The Horizon