Corporate spending on artificial intelligence is at an all-time high, yet C-suite executives admit that they're not seeing a lot of value from their investments yet, especially from generative AI.
Boston Consulting Group surveyed more than 1,800 C-level executives for its recent report, "
"The technology is at our fingertips," said Vlad Lukić, managing director and senior partner at Boston Consulting Group, in an American Banker podcast that will air April 1. "We're using it in our daily lives. Therefore, by default, we are expecting to see it in our workplace as well. It's available and the costs are dramatically decreasing."
But returns on this spending appear elusive: only 25% of respondents to the BCG survey said they are seeing significant value from AI investments. Among the executives Accenture surveyed, only 13% reported creating significant enterprise-level value.
To be sure, some banks are seeing returns.
"Any hour of productivity that we get from a developer multiplies itself by 12,000 – it's easy to do the math," Goldman Chief Information Officer Marco Argenti said in a recent interview.
"There are certainly some returns on investments, for example in identifying fraud, supporting agents in call centers or with assistants that can prep advisors, or scribing and transcription summarization tools," said Vik Sohoni, senior partner and global leader of McKinsey's banking digital analytics practice. "The other question is are these quantifiable, tangible and visible in the bottom line or are they ephemeral and a few minutes here and there? So far it's more the latter, with some exceptions like call centers."
Waiting for payoff
There are myriad reasons gen AI investments have yet to generate big payoffs.
One is that it's still early days. The gen AI boom arguably started in November 2022 when OpenAI made ChatGPT public, and some banks are still in test-and-learn mode.
Another is that banks must be cautious in their approach to generative AI, as the technology is subject to potential bias, hallucination and error.
"These methods are probabilistic and errors are inherently difficult to quantify, so banks as regulated entities need to be very prudent," Sohoni said. "The risk and compliance infrastructure requires maturing to keep pace with the nature, scale and pace of the machines' output."
An added challenge is that generative AI often works behind the scenes, said Michael Abbott, global banking lead at Accenture.
"I don't think gen AI is going to change banking, but it will rewire how banking is delivered," Abbott said. A lot of generative AI happens "below the tip of the iceberg" rather than something more visible, he said.
"Banks are taking a very thoughtful, risk-based approach to it, not unleashing it to their customers directly, focusing on mainly internal processes right now," he said.
Not having the right people lead gen AI projects is an issue for many companies, BCG's Lukić said.
"In some instances, they just appoint the wrong team to drive it, without accountability for results or without organizational position to actually drive impact," he said. For instance, some companies have well-established data science teams that the CEO asks to deploy generative AI. But deploying gen AI is different from building custom models or analysis.
"Data scientists are not anthropologists or people that know how to rethink business processes," Lukić said. "So they will usually take these models, they'll keep fine tuning them, increase the accuracy, but then they depend on the business process owners to deploy it. That's a very frequent failure point that we see."
A better way, he said, is to appoint business-process owners to drive these projects, with help from the data science and other technical teams.
Underlying management challenges like these is the need for top AI talent. "Banks themselves need to shape their value proposition to attract the level of talent that can implement and run these complex engines," Sohoni said. "Some have, while others will find it harder."
Many banks rely on vendors that themselves don't all have AI experts on staff, he said.
"These are highly complex capabilities and if you're a really talented AI engineer, your employer of choice might not be one of those vendors," Sohoni said. "So the vendor ecosystem has a ways to go to mature, too."
Another hurdle to achieving a return is the tendency to give employees tools that save them time but not tell them what to do with that freed-up time, Lukić said.
And some companies focus on the wrong metrics, he said. One example is streamlining a loan process without recognizing that because the loan committee only meets once every 10 days, the response rate to the consumer stays the same. If the bank focused on customer response rate, and perhaps modified the loan committee bottleneck, then the bank could see results.
Need to focus on revenue
Banks would be better off focusing on generating revenue than on trying to cut costs through advanced AI, Abbott said.
"The best banks I've seen are more worried about revenue-generation opportunities than they are worried about cost and the productivity gains," Abbott said. "I've seen banks all around the world customize their marketing truly on a one-to-one basis, with language based on the persona of the individuals and exactly what they want to do. I've seen people use it on the collections side to look at behavioral economics to figure out exactly the right words to say to ensure you're at the top of collections — that's a revenue opportunity."
Another example he cites is the use of generative AI in call centers. Quite a few banks use gen AI to summarize calls, saving agents the time and drudgery of this task, and they're seeing 15% to 20% productivity improvements, which means the agents can take more calls.
But the technology can also be used to analyze the customer file before a customer calls in, figure out everything the person might want to talk about and recommend next best actions "so that the call center rep can feel like it's a continuity of conversation and close the problem before it ever starts," Abbott said.
"That's a narrow aperture versus a wide aperture, and I think many people have narrowed the aperture too small, and by widening the opportunity you find much more creative solutions," he said.
