It’s barely 20 months since OpenAI launched ChatGPT, but in that short time we’ve seen a dramatic change in how financial services executives think about the future of their organizations. What has become clear is that the capabilities of this innovation are expanding more rapidly than most companies can take advantage of them.
This has huge implications. It means banks and insurers that are slow to respond are leaving value on the table, value which their more fleet-footed competitors are seizing as they transform their operating models, their customer experiences and more.
Ecosystem of AI agents
Rather than utilizing a large monolithic AI capability, or even a handful of models, most banks and insurance firms will likely deploy a large number of highly-focused generative AI agents. These are semi-autonomous software programs, each of which has special capabilities. They can perceive their environment, understand intention, and take complex actions to achieve their goals. They can learn, and either find or create the tools they need to carry out their tasks. Most will be connected to other special-purpose agents in ecosystems that accelerate workflows within business functions.
The majority of these AI agents will require minimal ongoing input from humans. Once they understand what is expected of them, they will use their capabilities, and communicate and collaborate with other agents in their network, to deliver this outcome.
One example is Devin. Its creator, Cognition Labs, calls it ‘the first AI software engineer’ because it can write code, build websites and develop software from simple prompts. The ability of Devin and its counterparts to learn from mistakes and continuously improve makes them far more sophisticated than the AI chatbots of years past.
Some AI agents will automate processes with little or no ongoing human input, while others will augment humans by responding to changing instructions. A priority right now for FS firms is to modularize these agents so they can be inserted quickly and easily, with little customization, into a variety of workstreams. There is still work to be done before this is a simple, routine procedure. Especially important is the establishment of effective controls that monitor the decisions which these agents take and, when necessary, can reveal the reasoning behind each decision.
Reimagining the business
The sudden arrival of generative AI agents presents FS executives with a novel challenge: to reimagine the future of their organization, a future in which talent and technology collaborate differently and more intensively than before. This has profound implications for the workforce, the nature of work and the culture of the organization. It adds greater complexity to managers’ traditional role of orchestrating their resources, going way beyond role restructuring and the new skills required, to include what a day at the office will be like when AI is embedded into much of the business.
Until recently, for example, banks’ loan underwriting and fulfilment process was heavily dependent on human beings. Now you can reconstruct this value chain using a collection of AI agents responsible for data collection, risk assessment, recommendation and loan fulfilment. A human advisor will ultimately make the final lending decision, but AI agents will handle much of the work, speed up the process and add value – allowing the advisor to review more loans faster.
Another example is the KYC and onboarding process, where AI agents can collect and analyze a wide variety of both structured and unstructured data (from media reports and financial results calls to videos and social media posts) to increase the rigor of background checks. Other agents in the network can take care of other steps in the process, such as drafting and sending personalized messages to potential customers.
Rules and values
A new challenge for banks and insurers is ensuring that AI agents not only meet their regulatory obligations but also share and reinforce the company’s values and goals. We worried less about the latter when our technology was restricted to the back office and did mostly what it was told. However, generative AI agents are rapidly becoming more powerful, autonomous and ubiquitous, making this an urgent priority.
There are also the technological challenges of enabling coordinated AI agents across increasingly scaled implementations – while ensuring the trust threshold of these agents can be met. Financial services executives will need to acquire a solid understanding of what generative AI agents are, how they work and where they should best be deployed. This will require the rare combination of tech knowledge and deep business process expertise that allows executives, for example, to understand how deficiencies in the firm’s data architecture limit the potential use cases for a particular AI agent. This capability will be invaluable as the pace of change accelerates and the time for decision-making contracts.
Changing the game
When generative AI landed, we were first astonished at its potency. Then we started thinking about where and how in the FS operating model we could use it. We are only at the dawn of the age of AI, and are still feeling our way as we try to fathom where it will deliver the greatest benefit and, in the process, transform key aspects of our business. We have some way to go before the mist clears, but that’s what makes this such a fascinating time to be in financial services.
Read the full article here
