The Commonwealth Bank of Australia (CBA) has embarked on a groundbreaking initiative to develop generative AI-powered chatbots for testing new products and messaging strategies. This endeavor, presented at the South by Southwest Sydney 2023 event, is in its preliminary phase. The primary objective is to evaluate the AI’s capability to simulate customer behavior and test the effectiveness of chatbots as an early-stage experimentation tool for rapid testing.
CBA’s Chief Decision Scientist, Dan Jermyn, explained that this initiative involves creating AI-based personas and generating various versions based on defined characteristics, such as age and industry. These personas are then used to create simulations that mimic how real people might interact in different scenarios. Importantly, these AI-based personas are not intended to replace the extensive market research conducted directly with customers. Instead, they serve as a way to experiment with new product ideas and messaging strategies in a safe and scalable manner.
Simulating customer responses in challenging situations
The bank has already developed various customer personas through this preliminary study, enabling the chatbots to express concerns, ask questions, and identify issues like real humans. CBA believes that this capability will be particularly useful for simulating customer responses in challenging situations where traditional face-to-face research is more difficult.
Furthermore, the bank sees the potential of generative AI for understanding customer needs during natural disasters and vulnerable situations. This includes simulating customer actions and needs during disasters and determining effective messaging for customers facing issues like potential scams or dealing with loss.
Jermyn highlighted the significance of AI transparency, as the foundational models play a crucial role in creating chatbots that closely mimic human behavior. He noted that slight changes in questions can result in very different behaviors.
Bank is exploring different forms of generative AI
One interesting aspect of this experiment is how average Australians are represented in large language models, which can differ from models constructed from the global population. This emphasizes the importance of CBA’s research to ensure AI is safely and effectively applied to Australia’s specific context.
This generative AI project is part of the CommBank Gen.ai studio, established in May, allowing CBA to adapt various models to its specific requirements. The bank is exploring different forms of generative AI and large language models, recognizing that different models may excel in particular use cases.
The adaptable and transparent use of AI is a fundamental principle guiding CBA’s AI initiatives, emphasizing the need for safety and scalability. This innovative approach reflects CBA’s commitment to bringing AI safely to scale while addressing specific Australian needs.