What businesses should prioritise in generative AI adoption

Banks should focus on integrating new technology into banking processes without forgetting that humans are at the center.

The key to a successful implementation of generative AI lies in balancing innovation with safety, ensuring that human elements remain central to any technological advancement, according to Na Zhou, Data and Analytics Lead Asia Pacific at Oliver Wyman.

Generative AI, known for its efficiency and business outcome improvements, stands as a powerful tool for banks. However, as Zhou pointed out, "These models complement traditional machine learning models but come with their own set of challenges and unfamiliar failure modes." 

Zhou advised a threefold approach for immediate priorities. First, banks should focus on gaining experience and building capabilities through the development of appropriate use cases. This approach not only provides clarity on risks and rewards but also aligns with top business priorities. 

“Number two, establish or upgrade your AI risk and controls framework with a focus on practical guardrails. And the third one is to get conditions ready. This includes getting the technology infrastructure ready, most banks need to start making some no-regret moves in that area,” she said.

Looking at the long-term strategic implementation, Zhou highlighted two crucial steps: Banks need to establish a clear view of AI ambition versus risk appetite and identify new business models' sweet spots, moving with careful consideration and speed.

Addressing the risks associated with generative AI, Zhou suggested a cautious 'test and learn' approach, especially in client-facing functions, to mitigate potential reputational and financial impacts. 

She discussed that measures for risk management fall into two main pillars: technology-focused strategies, like retrieval-augmented generation systems and anti-hallucination guardrails, and human-centered approaches, including governance processes and risk management for real-life consequences. This dual approach ensures both technological efficiency and human oversight, a balance crucial for the responsible deployment of AI models.

Zhou also addressed the skills gap challenge in implementing generative AI in banking, saying that a successful deployment demands a highly skilled workforce capable of understanding and leveraging this new technology.

“Indeed, the successful implementation of generative AI I demands a highly skilled workforce, who are capable of understanding and leveraging this new technology. Banks face challenges in upskilling their existing employees or recruiting new talent with expertise in AI or this new form of AI,” she said.

Zhou asserted that apart from training and education initiatives, operational experience with these new models is essential. Making this technology readily available to the workforce, fostering an AI adoption culture, and creating an environment that embraces technological change are pivotal steps.