Westpac collaborates with Kasisto to deploy Kai-GPT, a specialized AI model, ensuring accurate, secure customer and employee interactions.
Westpac, a leading bank in Sydney, is taking a significant leap in AI technology by integrating Kasisto’s Kai-GPT, a specialized AI model designed to provide precise and reliable banking services. This move addresses a crucial concern with general AI models, like ChatGPT, which are prone to “hallucinations” or generating incorrect information.
David Walker, Westpac’s Chief Technology Officer, emphasizes the need for accuracy and trust in banking services. The use of a public AI model, prone to inaccuracies, is not an option for a bank with over 12 million customers. Hence, Westpac is testing Kai-GPT, developed to deliver human-like responses while ensuring the information’s accuracy and relevance to the banking industry.
Kai-GPT, crafted by Kasisto, is trained exclusively on banking-related conversations and data. This focus drastically reduces the risk of generating irrelevant or incorrect information. Zor Gorelov, CEO of Kasisto, highlights the importance of accuracy, transparency, trust, and customizability in banking AI models. For Westpac, internal applications of Kai-GPT will empower front-line bankers, contact center agents, and mortgage workers.
One significant application is in mortgage lending, where Kai-GPT will assist in streamlining the application process. This integration aims to reduce back-and-forth communication with customers and enhance the overall efficiency of the loan process. The technology promises to provide more comprehensive and conversational assistance, a clear advantage over traditional chatbots.
Westpac’s approach includes layering the AI model: one layer is trained on data from multiple banks, and another layer on Westpac-specific information. This method ensures the generation of rich, accurate answers tailored to Westpac’s services and policies.
Peter Wannemacher, a digital banking analyst at Forrester, anticipates similar trends across the banking sector. However, he notes that banks will proceed cautiously, given the sensitive nature of financial information.
The initial implementation of Kai-GPT at Westpac will focus on mortgage operations, with plans to expand its application once proven effective and safe. This cautious, responsible approach aligns with Westpac’s ethical stance on AI usage, adhering to the principle of “do no harm.”
Walker envisions this technology as a game-changer, enhancing the quality of responses and customer interactions. Kai-GPT’s human-like responses bridge the gap between robotic chatbots and human conversation, potentially revolutionizing customer service in the banking industry.