SouthState Bank has embraced Microsoft’s Bard, a ChatGPT-like technology, significantly diverging from the cautious approach of other financial institutions.

Chris Nichols, the Director of Capital Markets, hailed Bard as “game-changing” at the American Banker’s Digital Banking Conference. The technology shines in text assimilation, summarizing vast documents, and aiding in email composition and marketing copy creation. The bank now finds it invaluable for summarizing policy documents and regulatory material.

SouthState Bank has fortified Bard’s capabilities by exclusively training it on bank documents, steering clear from integrating customer data. This method ensures a secure environment, free from potential data leakages that concern many in the banking sector. The tool acts as a robust search engine, offering complete answers instead of mere links, and going a step further by intelligently linking ideas to form coherent thought in text. This novel approach has allowed employees to sift through lengthy text threads, summarize discussions, and pinpoint action items swiftly.

The implementation has yielded tangible productivity gains. About 2,000 of SouthState Bank’s 5,000 employees are actively using Bard to navigate through the bank’s extensive commercial loan policy and branch policies. The technology has trimmed down the time taken to find correct answers from 12 to 15 minutes to mere seconds. This sharp reduction in time spent on information retrieval translates to a five to eight-fold boost in productivity. For instance, new tellers can now easily inquire about procedures such as reissuing an ATM card, receiving precise guidance swiftly.

However, the journey is not without challenges. The accuracy issue, particularly the occasional “hallucinations” where the system fabricates answers, necessitates human oversight. Moreover, updates to policies require careful handling to ensure Bard grasps the new information while discarding the outdated data. The technology’s deployment cost, at $50,000, and its monthly run cost of $30,000, is seen as justified given the substantial productivity enhancement. SouthState Bank’s venture into generative AI underscores the transformative potential of large language models in the banking industry, setting a compelling precedent for others to follow.