• Digital Economy Dispatches
  • Posts
  • Digital Economy Dispatch #180 -- The AI-at-Scale Investment Dilemma: Managing Risk and Delivering Value

Digital Economy Dispatch #180 -- The AI-at-Scale Investment Dilemma: Managing Risk and Delivering Value

Digital Economy Dispatch #180 -- The AI-at-Scale Investment Dilemma: Managing Risk and Delivering Value
21st April 2024

In 2000, Sterling Software was purchased by Computer Associates (CA) for more than $3Billion. At the time I was Vice President of Research and Development at Sterling Software, and the news of the purchase by CA came as quite a shock. Within days the management team made the trip to meet the new owners in Long Island, NY, where we presented our plans. I made my pitch about the research we had underway, the rapidly evolving technology landscape, and the new product plans that we intended to pursue. Then I waited for a response.

The result was a heated debate. The new owners clearly did not share my enthusiasm for investing in new technology. They were focused (some would say “obsessed”) on driving down costs and pursuing a very cautious approach to adding product capabilities ahead of demonstrable market demand. After several hours it was clear we were on different paths. They followed theirs, while I left within a few weeks to join a new digital startup in Silicon Valley.

The experience was painful. Yet, it also taught me a great deal about the dilemma we face as we design digital strategies that address the speed of new technology adoption, the challenges of balancing innovation with fiscal responsibility, and the way risk is perceived by different individuals, teams, and organizations. A drama that we see being replayed over and over again as we look at the adoption of the latest wave of AI technology. There is a great deal to be learned by considering how this struggle is being played out. And one of the most interesting domains to observe this is the financial services sector.

The Digital Transformation of Financial Services

Generative AI is rapidly transforming the landscape of financial services, promising unprecedented opportunities for efficiency, personalization, and innovation. From streamlining customer service to revolutionizing software development, we are already beginning to see the impact.

However, realizing these benefits requires navigating complex challenges and organizational dynamics where ambition and caution clash as financial institutions revise their future digital strategies to embrace these latest advances. Investment in the new waves of AI technology must be measured against the operational constraints of running a large, complex organization. How financial institutions manage the high expectations for generative AI adoption within the realities of large-scale digital disruption offers an important lesson for everyone.

There is no doubt that Generative AI offers substantial benefits to the financial services sector. McKinsey emphasizes several key areas where it is extending and augmenting more established uses of AI.  Most visibly, virtual assistants are being powered by generative AI to streamline customer service by guiding loan officers and providing pre-approved document templates. This frees up human employees to focus on more complex interactions. However, more broadly generative AI models act as intelligent assistants, empowering employees by processing vast amounts of data. This includes AI summarizing regulations, generating research reports, or creating instruction manuals – all this translates to valuable time saved for human employees who can then focus on strategic tasks.

In terms of customer benefits, increasingly, generative AI is being applied to generate personalized content in real-time, allowing financial institutions to tailor marketing and sales materials to individual customer profiles. This level of personalization is being used to significantly enhance customer experiences.

Finally, IT operations is an area where generative AI is having a major impact across many sectors. Generative AI revolutionizes software development, a core capability for all major financial institutions. Code-writing AI assistants can translate legacy code and support developers with debugging and testing, ultimately accelerating the software delivery process.

Meeting the High Expectations for AI

All of these potential uses and early successes of generative AI are leading to high expectations for its widespread deployment across financial institutions. For example, Accenture’s analysis of the potential adoption of AI across different banking roles indicates that 73% of the time spent by US bank employees has a high potential to be impacted by generative AI—39% by automation and 34% by augmentation. Furthermore, its potential reaches virtually every part of banks and insurance companies, from the C-suite to the front lines of service and in every part of the value chain.

However, experience with digital technology adoption over several decades has revealed that the optimistic forecasts from technology proponents often face challenges when confronted with the reality of effecting significant shifts in organizations operating within rigid governance structures, relying on aging legacy technology, and navigating complex regulatory environments. All these factors must be addressed in defining an appropriate strategy to leverage AI effectively.

An Illustrative Scenario

Based on these opportunities and challenges, formulating an appropriate AI strategy in financial organizations is anything but straightforward. This is aptly illustrated in the hypothetical scenario described by Thomas Davenport and George Westerman. They highlight the issues using a fictitious scenario to portray the clash between the ambitions of technology leaders in a financial services organization and the budgeting and governance constraints faced by finance and risk management executives. It is a dramatization of a real situation that is being experienced today in the board rooms of many financial services institutions.

In Davenport and Westerman's example, the CEO of a fictional bank contemplates a substantial investment in AI as the head of AI innovation advocates for a large-scale plan to transition it into an AI-first bank. This proposal entails significant staff restructuring, with AI assuming most customer interactions while human staff focus on complex issues and high-value customers.

In contrast, the CFO exhibits caution regarding the proposal. They express concerns about losing the human touch, potentially alienating customers and employees, and the substantial cost and uncertainty of project success, drawing from past experiences with cost overruns in major digital transformation endeavours over the preceding two decades.

In this familiar scenario, the CEO finds themselves caught between two opposing viewpoints. The potential benefits of AI clash with the importance of maintaining a human connection with customers and a more measured approach to AI adoption. The decision regarding the extent and pace of AI investment will determine whether the bank remains a traditional institution gradually digitizing to enhance operational efficiency or undergoes a significant transformation embracing AI-at-Scale to redefine itself for a new era of financial service offerings.

While this illustration is fictitious, its core characteristics reflect many of the issues confronting the financial sector. A report from UK Finance and global management consultancy Oliver Wyman examining AI utilization within financial services corroborates the opportunities and associated risks for the sector. This comprehensive analysis draws insights from a snapshot survey encompassing 23 companies that includes multinational entities to mid-sized banks and non-banking financial services firms.

The findings reveal that a significant majority (70 per cent) of surveyed firms are currently in the pilot phase for generative AI, particularly focusing on 'co-pilot' type tools aimed at enhancing employee efficiency in content production. However, it is anticipated that the realization of returns on investment for more sophisticated applications will typically span between three to five years.

Interestingly, three-quarters (75 per cent) of these financial services firms express confidence in the benefits to be derived from generative AI, with the primary advantages foreseen in productivity enhancement and operational streamlining rather than in customer-facing or revenue-oriented contexts. Notably, a trial conducted by Marsh McLennan involving a generative AI assistant yielded positive feedback, with 94 per cent of users reporting increased productivity.

Despite this confidence, many important concerns are also highlighted in this survey. The report states that 95 per cent of the surveyed firms are investing in actively factoring AI risks into their control frameworks, and a significant proportion (60 per cent) have already implemented measures to mitigate the risks associated specifically with generative AI.

Lastly, a notable consensus emerges among four out of five financial services firms (80 per cent) who emphasize the importance of collaborating with regulators. Such collaboration is seen as instrumental in promoting best practices in AI deployment and in fostering the development of an internationally aligned regulatory framework.

Challenges and the Road Ahead

While AI offers a promising future for financial services, there are hurdles to overcome. Just as in other sectors, issues such as cybercrime and ensuring customer data privacy and security within chat interfaces is paramount. Furthermore, training AI models to understand the nuances of financial services language is another challenge. Finally, fostering customer adoption through education user-centric service design is crucial for successful AI integration.

In addition to these common challenges, for financial services success appropriate application of regulations plays a critical role in fostering responsible AI adoption. This is highlighted in the Bank of England’s investigations in AI which emphasize the need for a balanced regulatory framework that supports innovation while mitigating potential risks to consumers, firms, and financial stability. They conclude that a key step involves clarifying how existing legal requirements apply to AI usage. Additionally, new industry standards and codes of conduct are necessary which instil trust in users by ensuring AI systems adhere to widely accepted ethical norms appropriate to financial services.

As seen in many domains, AI is fundamentally reshaping the financial service sector. From fraud detection and credit risk management to personalized customer service and efficient operations, AI's impact is undeniable. As the industry embraces the latest generative AI, the key will be to balance the need for fast paced innovation in products and services with the responsible adoption of AI technology in large established organizations. The pressure on digital leaders to get this balance right is intensifying.