McKinsey & Company has now spent two years assisting bank clients write software to make money using artificial intelligence, says Larry Lerner, a partner at the consulting giant who leads its global banking and securities analytics practice. During that time, McKinsey has developed an elaborate rubric to help chief executives identify problems most ripe for AI solutions, and which ones are likely to pay off first.
In conversation with Coins2Day Lerner broke down a report McKinsey published today, which he calls a guide for how banks can implement artificial intelligence.
“There’s lots of institutions that are taking the thousand flowers bloom approach, which is not leading to broad-based impact,” says Lerner, who has been a partner in the D.C. Area for the past seven years. “Those that are picking their battles carefully and going very deep in a way that they can reimagine full domains or full workflows, are really the ones that are getting the early momentum.”
McKinsey published its first big AI report last December. Since then, the consulting firm has joined the small army of consultants working with banks and other financial firms to help them best capitalize on the promises of the technology. According to Lerner, banks that have a successful track record using AI have one thing in common: focus.
To help banks frame the problem, the report breaks down a typical institution into four domains and six subdomains, though the examples are not intended to be comprehensive:
While a technically simple experiment that frees up employees to do other tasks may be tempting, Lerner says bank executives should be mindful of the difference between such an efficiency play, which may take years to generate revenue, and something that might take a larger investment but almost instantly hit the bottom line.
“AI and Gen AI can drive revenue growth, cost savings, and cost avoidance,” says Lerner. “But if the first place you’re going to go is an efficiency play, where you can’t really measure what somebody can do with that extra 45 minutes a day, that is not going to truly flow down to bottom line results.” Together, these subdomains can add as much as 80% of total incremental value from using AI, McKinsey estimates.
The flip side is the trap into which banks fall when they fail to properly implement AI: Choosing an expensive, high impact subdomain and then understaffing it, which can be as counter-productive as choosing a low-impact solution and properly staffing it. Banks that fail to properly staff their project, or as golf-fan Lerner puts it, “under-club” it, are equally doomed to fail.
“The institutions that are failing to scale today are largely doing so because of an under-clubbing on the execution muscle that they have to build in order to bring this into production,” says Lerner.
Another warning Lerner shared with Coins2Day, not covered in the report, is the temptation to decentralize. While banks and many other large institutions have been pushing more and more authority to the edges of their organizations—empowering the people in the trenches to solve their own problems—Lerner says this process is no longer beneficial when developing complex artificial intelligence solutions.
“The leaders are centralizing more,” says Lerner. “Which is allowing them to make bolder decisions. The laggards are taking a much more expansive approach to opportunities and proofs-of-concept, which is not allowing them to really get that bottom line impact that they were looking for.”