The Two/Three-Step-Forward, One-Step-Back Journey of AI

“I'm not worried about AI replacing me. Can AI sit in a fleece vest at a bar all day and do nothing?” –John Mulaney

Gonzo nation, the focus on AI in banking is everywhere. You cannot go to a management or board planning session without AI emerging as a topic before the morning coffee is served. And for good reason—multiple studies show that AI spending in the United States will top $200 billion in 2026. Chase alone will invest over $2 billion.

The early use cases are compelling. Improvements in document management, task automation, customer service, fraud detection, compliance and treasury, among others, show that we are in the early stages of a fundamental transformation of banking.

That said, the journey is not, nor will it ever be, on a straight road with no bumps. Consider some recent news:

• Ford Motor Company recently re-hired 350 veteran engineers after its AI-driven quality systems failed to deliver the results it expected (score one for the gray-hairs!).
• Klarna, the Swedish buy now, pay later company, is hiring back up to 700 agents, at least partially reversing its year-old artificial intelligence-first strategy. The quick and dirty: Handle time on calls went down, but repeat calls increased 25% because customers didn’t see their issue as being resolved.
• BCG research showed that poor data quality has derailed 40% of AI projects.
• A recent Forbes study reported that 58% of employees surveyed said they are indeed saving up to five hours a week using AI tools (that would amount to about 1.4 million employees just in banks and credit unions). However:
  • Another 19% reported having used AI at work without any time benefit.
  • 46% said they think that the time they save using AI belongs to them and not the company. Yet most also said they would still spend at least half of their reclaimed time on work-related activities.
  • Half said that the newfound work time they did give back to the company would be focused on doing existing work better, not necessarily doing more work.
  • Hmmm. That five hours is starting to look like it nets to more like 2-4.

So, fundamental change is happening because of new technology. With slips, slides, do-overs, cancelled initiatives, and a decent amount of wasted time.

But here’s the thing. We have been here before. Maybe we didn’t face as fundamental a change as we are with AI, but we have been here before. And we learned some things.

• Desktop tools—Starting about 40 years ago, desktop tools, in particular Excel and PowerPoint, fundamentally changed the way employees analyzed data and presented it.
• Database and reporting tools—About 20 years ago, these tools allowed bankers to store, normalize, and analyze large amounts of data in a way they could not before.
• CRM systems—Starting about 10 years ago, CRM systems became mainstream and brought new dimensions to customer relationship management and marketing that were just unavailable up to that point.

These tools unquestionably changed things for the better. But they also left plenty of litter on the road:

• Literally thousands of spreadsheets at every bank that people seldom or never use (Think I’m wrong? Go count on your servers now. There is a huge graveyard.)
• Multiple reports created by different people that show exactly the same thing
• Decisions made with bad information due to poor database integrity and reporting errors
• Conflicting reports or conclusions that required resolution (remember those “where did this come from” meetings?)
• Marketing and sales initiatives that produced no real results
• Too many employees using tools to become more efficient but in the end just wasting time (Some of this can be written off to learning, but not the majority.)

Now we face the explosion of AI tools that go well beyond the three mentioned here. Now we are redesigning processes, redefining learning, rethinking customer experience, and stepping up data management and analysis. It will make bank performance better.

But as before, there will be litter. Smart banks will take lessons of the past to minimize it. They will:

• Focus on corporate AI initiatives that create real, measurable ROI. We have talked in the past about the difference between corporate AI initiatives and employee use of AI to produce individual results. Where do you think the most time will be wasted?
• Be very clear with employees about what AI tools they can use, private and public versions. Hopefully, most of you already have published corporate policies.
• Be transparent, and communicate the prioritized initiatives to everyone.
• Make sure you track, manage, and take advantage of vendor AI initiatives. Every one of your vendors is embedding AI into the systems you bought and use. Don’t duplicate unless their efforts fail.
• Track, communicate, and standardize individual AI productivity tools created by employees. For example, if one employee comes up with a good idea to manage his/her email inbox, and it works, don’t let dozens of other people create something else just like it. Make it the standard. (Note: this happened at Cornerstone. I speak from experience here. And no, I did not create it.)
• Enforce the use of standard, managed, normalized data. AI will depend on good data every bit as much as previous systems did.

Mark Twain famously said, “History does not repeat itself, but it rhymes.” Smart banks will take full advantage of AI as it grows and matures, but they will also use past experiences to guide success and minimize litter.

 Terence Roche is a founding partner at Cornerstone Advisors. Follow him on LinkedIn and X.