GENERATIVE AI

GenAI is a topic that keeps coming up at our forums. Like RPA before it, the question ripples throughout our groups: What are you doing about it? And, like RPA, the answer for the first three years is: we’re closely monitoring the situation. Then, for roughly two years, “we are testing a few bots.” Two years hence, “The technology didn’t live up to its promise and didn’t result in meaningful scale economies.”

(Gen)AI is equally mysterious in many ways. Thankfully, I had the opportunity to listen to Larry Lerner, Partner and Global Lead Banking and Securities Analytics at McKinsey. He enlightened me somewhat, and I want to share some of his wisdom with you.

The context offered by McKinsey for this new technology is: “A natural evolution of analytical AI, addressing a new set of challenges. While analytical AI analyzes patterns it can discern within data, Generative AI creates something entirely new from what it can see. Generative AI algorithms can create new content in conjunction with people to enhance people’s output. It offers a new engagement level between humans and technology that can leverage both more effectively. While RPA’s main strength entails data aggregation, reporting, and pattern identification (e.g., generating credit spreadsheets), GenAI goes beyond this to create new ideas, potentially customize 1:1 communications, and even act as a virtual assistant to employees by automating tasks they currently do themselves, as well as do the same for clients.

McKinsey estimates that 90% of the expected GenAI impact will be concentrated in four business domains: software engineering (including code development), customer service operations, marketing and sales, and risk management and legal.

The technology is taking root at several large companies, from JPMorgan and Morgan Stanley to Goldman Sachs and Moody’s. Use cases range from customer service functions (think Voice Response Units on steroids and more effective chatbots) to automatic summaries of meetings, client interviews, etc., and generating drafts for reporting and future emails. In addition, knowledge assistants powered by GenAI house huge amounts of data and help financial advisors quickly track down information, which can greatly benefit consultants, attorneys, wealth advisors, and more.

Like its predecessor, GenAI’s promise is great but vague. Very few implementations have a quantitative goal associated with them, and many are committed to trying the technology on the planning horizon.

Again, like its RPA predecessor, GenAI encounters numerous user concerns, from intimidation and complexity to a real fear of losing one’s job. The hardest thing about GenAI is change management. We have found that this is a major impediment to technology adoption, which is particularly problematic as technology is rapidly changing. Success requires intense strategic focus and meticulous workflow tracking, both challenging and atypical of our industry.

Meanwhile, as we contemplate what the new technology can and should do, others are embedding it in their products so that it is both available and sometimes used by customers who bought the primary product. ChatGPT is an example of such technology, and the new iPhone offers the next generation of such adoption.

I believe this is a great opportunity for all of us to personally try GenAI and get a first-hand experience of what it can do its shortcomings, and the related aggregation that all new technology carries with it. I have ChatGPT on my phone and use it rarely, primarily to remind myself of its capability (ask for a steak recipe, for example).

As an observer from the sidelines, I am humbled by experts like Mr. Lerner, who know so much about the space. At the same time, the pace of changes seems so intense that, to a degree, we are all learning simultaneously. Community and even much larger banks are unlikely to outpace their trillion-dollar-plus brethren regarding investment, use cases, or adoption. In some instances, early adoption benefits far outweigh the risks (e.g., ApplePay participation).In others, such as this one, learning from others seems to be the way to go.

One preliminary step that will always serve you well is mapping out your major workflows—inefficiencies set from workflow imperfections and loops and poor data. GenAI will be most impactful when your workflows have been cleaned up, and data hygiene rules your databases. The benefits from such efforts will transcend any technology and are worth pursuing ASAP.