According to Forbes, McKinsey & Company’s new report reveals banks are entering the era of “agentic AI” – autonomous systems that can plan and execute multi-step workflows with minimal human intervention. This shift from broad experimentation to precision deployment could reduce bank unit costs by 15 to 20%, but also threatens to erode up to $170 billion in global profit pools by 2030 if business models aren’t adapted. Anthropic’s Jonathan Pelosi identifies 2026 as the psychological tipping point, noting AI accuracy has improved from 80% to 99% in factual reporting. Meanwhile, Amazon Web Services’ Scott Mullins observes banks are moving beyond “AI tourism” toward specific business outcomes, focusing on modernizing decades-old legacy systems and automating compliance workflows.
The Unsexy Revolution
Here’s the thing: everyone’s talking about chatbots, but the real money is in the boring stuff. Banks are finally realizing that flashy customer-facing AI gets headlines, but operational efficiency gets results. We’re talking about modernizing COBOL systems that nobody even knows how to code anymore, automating compliance workflows, and reconciling ledgers without human intervention.
Basically, banks are repairing the plane while flying it. They can’t shut down core systems to upgrade them, so they’re integrating AI agents directly into live environments. It’s not glamorous work, but it’s where the 15-20% cost savings actually materialize. And honestly, this is where the real industrial-grade computing power comes into play – the kind of reliable hardware that companies like Industrial Monitor Direct specialize in for mission-critical operations.
The Threat From Shopping Agents
Now here’s where it gets scary for banks. For decades, they’ve benefited from customer inertia – people just don’t bother switching accounts to chase slightly better interest rates. But agentic AI is about to change that completely.
Imagine AI “shopping agents” that constantly monitor rates and automatically move your money to the best accounts. McKinsey estimates that if just 5-10% of checking balances migrated this way, deposit profits could decline by 20% or more. Suddenly banks aren’t just competing with each other – they’re competing with algorithms managing their customers’ financial lives. That’s a fundamentally different game.
The Human Sandwich Approach
So how do banks deploy this powerful technology without creating catastrophic errors? They’re adopting what Pelosi calls the “sandwich” approach – AI agents handle the heavy lifting, but humans define the goals and validate the outputs.
Mullins from AWS suggests a “golf bag” strategy too: using different AI models for different tasks rather than relying on one vendor. This makes sense when you think about it – you wouldn’t use the same club for every shot in golf, so why use the same AI model for every banking workflow? The key is maintaining human oversight while letting AI do the grunt work.
We’re seeing a maturity in thinking here that was completely absent during the initial AI hype cycle. Banks are finally asking the right question: not “What cool AI thing can we do?” but “What specific business problem are we solving?” And that shift from wow to how might just determine which banks survive the coming algorithmic competition.
