Brett King: Bankers Have Five Years to Rebuild for Agentic AI, Or Perish
By Steve Cocheo, Senior Executive Editor at The Financial Brand
Simple Subscribe
Subscribe Now!
Banking futurist Brett King says there are two types of bankers today.
• First type: Those experimenting with AI, figuring out how to apply it in their daily life and in their business.
• Second type: Those who are going to be replaced by the first.
Which are you going to be?
Already, U.S. job openings are off by 30% since the advent of the ChatGPT era, King points out. And the trend is only going to accelerate.
Why AI is different. King acknowledges that many of his banking and corporate contacts disagree with his bleak outlook. But he’s adamant.
“They argue that every time we’ve introduced a new technology, we have always replaced jobs that we’ve destroyed with new jobs,” says King.
“But we’ve never had a technology like AI that can be applied from the white-collar level to the blue-collar, across every sector simultaneously,” he argues. “So the potential for shocks to the economy in terms of human capital is unique.”
The stakes: King predicts that as agentic AI is wound into the business, everything about banking will change, from how banks and customers find each other, to how products are fitted to customers, to how the business is regulated and how institutions ensure compliance.
Time is short: Adoption of agentic AI isn’t as simple as downloading software. For all but the largest, most adept traditional banks, says King, it will take five years to put the basics of agentic AI in place.
Need to Know:
- By 2035, a significant share of the world’s money movement will be accomplished by agentic AI using payment rails.
- By 2050, 35% of all global corporate enterprises will be run by agentic AI, projects King.
- Getting ready will require rethinking of processes, breaking through data silos, and revamping how banks are organized.
- Traditional performance metrics will go out the window. New yardsticks must be crafted to account for AI doing the heavy lifting.
Agentic Banking is Embedded Banking Writ Large. Very Large.
Figuring how to fit into banking’s future begins with understanding what agentic AI means for the business. For years King has advocated for embedding banking into other processes, so that the “bank” becomes an invisible function. Agentic AI banking is the next stage in that evolution, he says.
By definition, agentic AI in banking has agency, so on both sides of a transaction AI will be interacting with AI.
While there will be some human intervention, approving next steps and transactions, increasingly agentic AI will handle more and more things directly. Some iterations from King:
• A basic case: A shopper tells their agent to find a particular item. Having found it, the agent seeks approval. Then payment comes in — if there isn’t enough money in ready accounts, the agent explores credit options.
This happens in seconds.
• A more advanced case: Today, voice assistants and related technologies enable AI to match customer needs to the right off-the-rack products that come close to meeting those needs. King says agentic AI will go further, taking all the raw materials the bank can muster and stitching together a tailored solution.
“This is true generative finance, where essentially we’re taking the core utility of the bank and putting a wrapper around an agentic AI solution and delivering that,” according to King.
Both cases involve the “engagement layer” of agentic AI — where the bank meets the customer.
Dig deeper:
How Agentic Banking Will Transform Internal Operations
But much more will have to happen behind the scenes, and all of it driven by AI, not humans.
King says this includes such functions as:
• The agentic handler — “It’s like a traffic cop, making sure there’s consistency, tracking rogue behavior by other agents.”
• Autonomous “know your customer” (KYC) and related compliance agents.
• Dynamic credit scoring agents.
• Agents to track fraud being attempted at scale by outside agents.
• Agents to detect hallucinations and other anomalies.
Speed will drive change. As more happens inside the black box of AI at hyperspeed, how even the basics get accomplished will change — it will have to.
“Today there’s people watching people watching people,” says King. “But we’re making the infrastructure far more efficient.”
King things AI will significantly outperform human operators in fraud interdiction, with fewer false positives and errors.
Read more: Five Real-World AI Applications That Will Boost Your Bank’s Operations
How Agentic AI Will Be Trained
“It’s not like this technology is going to come in and try to figure out from scratch how to do things,” says King. “We are teaching it how to do the processes that humans are doing. It will just be doing it far more efficiently.”
That’s for starters, and a job, for now, for human bankers. But agentic AI is designed to look for new connections and opportunities.
“The agents are going to see that ‘This is the data I have, this is the request that’s being made, this is the output of the decision that must be made. How do I get from Point A to Point B?’,” says King. “And then it’s going to find more efficient ways of doing that.”
Read more: We’re Speeding Past the Tipping Point of Gen AI in Banking
Rebuilding the Guts of Your Bank
Much of what King has to say concerns what he calls “cultural agility” for banks. But that must go hand in hand with technical agility, he says.
A hard reality. Your bank isn’t going to be able to run agentic AI on traditional core processing systems. Instead, banks will have to adopt a layer on top of their core — sometimes called a “sidecar” core – something most banks lack today.
Agentic AI requires such processing and “it’s not like you can go to a current core vendor and say, ‘Can we buy the agentic AI plug-in?’,” says King. “It doesn’t really work like that.”
Likewise, the data that feeds the AI must be made available differently. King notes that the importance of establishing “data lakes” has been discussed for years. But now data lakes, vast repositories of un-siloed customer information, are essential because they enable hypersonalization.
Read next: AI-First Nubank is Seeking a U.S. Charter, and Domestic Players Should be Worried
Reimagining Regulation as Banking Becomes Agentic Banking
“In the future, regulation will be in the code,” says King. “There’s no way around it. You can’t regulate a code-based system with humans trying to figure out what the black box is doing.”
While other countries have begun to implement elements of this approach, the U.S. doesn’t have anything in place yet. “You need regulators who are prepared to put in ethical guardrails and principles that AI can align to,” says King.
Down the road, AI-driven regulation may need to be designed to adjust on the fly just as agentic AI will.
Read next: Brett King: Why Branches Will Never Be the Center of Banking Again
