- McKinsey survey finds Taiwanese banks adopting Generative AI (GenAI) with majority engaged in testing or implementation.
- Anticipation of large-scale AI applications within one to three years, with warnings on integration costs.
- Common GenAI applications include operations, marketing, IT development, compliance, risk management, and talent.
- McKinsey underscores the need for broader capabilities beyond Large Language Models (LLMs) in GenAI adoption.
- Challenges include legacy systems, organizational restructuring, and talent acquisition.
- Regulatory guidelines for AI in finance expected from Taiwan’s FSC in June 2024.
- Urgent need for strategic AI budget planning at the corporate level, considering external challenges and internal integration hurdles.
Main AI News:
A study conducted by consulting giant McKinsey & Company reveals that the majority of Taiwanese financial institutions have initiated the implementation or testing of Generative AI (GenAI).
The finance sector anticipates significant AI deployment within one to three years. However, experts caution that AI utilization entails more than merely selecting the optimal Large Language Model (LLM), stressing the importance of considering the associated costs of system integration.
GenAI Applications in Finance
McKinsey’s inaugural examination of GenAI utilization in Taiwan’s banking domain involved senior figures from 15 banks, collectively responsible for approximately 70% of the industry’s total net income.
The findings indicate that 75% of banks have commenced GenAI adoption through diverse pilot initiatives, yet fewer than 10% have transitioned to widespread implementation.
Common application scenarios encompass operational processes, service enhancements, marketing strategies, IT advancements, regulatory compliance audits, risk mitigation, and workforce management. The predominant use case lies in operational streamlining and service enhancement, including the deployment of internal knowledge virtual assistants and basic Q&A applications for staff.
Charles Tan, a McKinsey partner, underscores the pressing demand for IT professionals within the banking sector. With legacy programming languages like Cobol facing a shortage of maintainers, the advent of GenAI tools capable of supporting coding tasks is accelerating processes that previously demanded five to ten years for language conversions.
LLMs: A Comprehensive Solution
McKinsey senior partner Violet Chung stresses that GenAI encompasses a suite of capabilities beyond just LLMs. She emphasizes that while some organizations prioritize identifying optimal solutions, they often overlook other crucial considerations. Chung analogizes LLMs to the human heart, which relies on a network of veins and systems for effective operation.
Chung underscores that purchasing solutions represents merely the initial phase. Irrespective of the provider, the standard package entails organizational restructuring, localization efforts, and employee training. Bridging the gap between acquiring services and achieving impactful application demands concerted effort.
Tan, conversely, notes that many commercial software codes are immutable, prompting firms capable of model fine-tuning to leverage off-the-shelf LLMs and customize them as necessary. Open-source models, offering greater customization flexibility, are gaining traction globally.
Addressing the necessity of additional AI infrastructure procurement, Tan clarifies that for every dollar allocated to specific GenAI software and hardware resources, an additional five dollars may be required for associated system integration. This encompasses data management, IT architecture enhancements, process modifications, and staff training.
The Race for AI Integration
Victor Kuan, a senior advisor at McKinsey Asia, highlights that GenAI has evolved into a “must-win battle” for numerous CEOs, driven by concerns that competitors might launch innovative services ahead and lure away customers. Nonetheless, he underscores the imperative for AI budget planning to be underpinned by a strategic framework from corporate headquarters or financial conglomerates, rather than ad hoc procurement by individual departments or treating it as a generic technology initiative.
The survey also identifies external challenges such as regulatory constraints, technological advancements, and talent shortages as significant impediments for banks.
In 2023, Taiwan’s Financial Supervisory Commission (FSC) unveiled a draft guideline for AI utilization in the finance sector. Following a public feedback period, the formal directives are slated for release in June 2024.
While these guidelines will be advisory rather than legally binding, certain firms express apprehension regarding potential influence from the FSC via administrative inspections. The McKinsey team advocates for swifter responses from regulatory authorities and clearer delineations of scope.
Conclusion:
The Taiwanese banking sector’s embrace of Generative AI presents both opportunities and challenges. While the adoption of advanced technologies like GenAI promises efficiency and innovation, the road to implementation is fraught with complexities, including integration costs, talent acquisition, and regulatory compliance. Companies must approach AI adoption strategically, focusing not only on selecting optimal solutions but also on broader organizational restructuring and talent development to fully realize the potential benefits of GenAI.
