# AI in Banking: Banks Are on Track, But Two Gaps Could Limit the Return
_Published: 2026-09-17T10:00:00.000-04:00_

Banks aren't behind on AI. Financial return comes later, and getting there depends on your people and your data as much as the technology.

_The financial payoff from AI takes time. Here's what banks should actually watch on the way to real returns._

Banks have done the groundwork on AI. In [nCino's AI in Banking Benchmark](https://www.ncino.com/ai-in-banking-benchmark), 91% of executives report having an AI strategy, and 71% have set KPIs to track it. At first glance, though, **one finding reads like a warning.**

Among the KPIs banks use to track AI, the two tied most directly to financial return — reduction in operating costs and increased revenue — sit at the **very bottom of the list**, as seen below. Plus, more than four in five (81%) say their organization has prioritized adoption over ROI.

The easy conclusion is that banks have rushed into AI and can't yet show what it's worth, **but there's a better way to read it. **Think about it the chart as a timeline, not a scoreboard. Financial return KPIs sit at the bottom because **they're further on the horizon**, not because banks are failing to reach them.

That's how [Danielle Pearson](https://www.linkedin.com/in/danielleatncino/) sees it. Our market analyst has spent years helping banks measure the return on their technology investments, so we asked her to walk us through what the survey found.

"**Adoption always precedes value,**" she said. "If you look at those priorities as a timeline rather than a hierarchy, banks are being pretty level-headed about where they are on the journey to ROI."

A bank can buy the best software in the world, but until people use it, it returns nothing. By that logic, banks focused on getting AI adopted are **right where they should be. **They've done the groundwork, too.

So the challenge ahead isn't the missing financial metric. That will come. The challenges to watch are two things sitting just beneath a healthy picture, and neither one is something AI can solve on its own.

## Why Return Comes Last

The path from adoption to financial return runs in a predictable order. Adoption comes first, then come specific operational returns: documents summarized in seconds, loans moving through the pipeline faster, more accounts opened in a day. Those wins accumulate, and **over time they compound into the financial returns**, like lower costs and new revenue, that show up last.

Seen that way, the KPIs banks report today make perfect sense. Productivity and workflow automation lead the list, followed by customer experience (44%) and operational efficiency (36%). Those are the operational milestones you would expect a bank to hit first, and they're the right things to be measuring at this stage.

Early wins are already real. Within three months of bringing agentic AI into its commercial lending, [**ConnectOne Bank cut document-search time by 98%**](https://www.americanbanker.com/news/how-connectone-bank-uses-ai-to-save-time-on-admin-work?blaid=8802861)**,** **from about 20 minutes to roughly 30 seconds**, and an AI agent reduced the time to update client relationships in documents by 60%, as reported by American Banker.

"The last thing you get is revenue realization," Pearson noted. "The first things you get are measurable returns on very specific things, and those compound and aggregate into the financial returns over time."

Those wins are encouraging, but lasting return takes more than technology; **it takes a workforce ready for the shift and data a bank can trust.**

## AI Still Needs Your People

Most banks expect a future where [people and AI work side by side — a dual workforce.](https://www.ncino.com/blog/ai-in-banking-dual-workforce-already-here) According to the benchmark, more than eight in 10 (84%) are deploying AI across the organization, and 89% expect their institution to be a mix of humans and AI agents within five years.

That future goes beyond just purchasing new technology. It’s something that is built, and the raw materials are a bank's people. Banks that will gain a real advantage now are the ones that bring their people with them.

**Reskilling and training is one step to prioritize now**, and it's already the most common one banks are taking, according to the benchmark. But only 55% are actively reskilling, and just 33% are hiring for AI experience. These numbers trail the near-universal expectation of a dual workforce, but they show early momentum.

People matter here because **_people make the technology work_**. AI can't, and shouldn’t, function without them. Judgment, originality, and relationships are what AI can't replicate. And banking, where trust and relationships carry the business, is about as human as industries get.

So the fear that AI will replace people gets the story backward.

"AI is never going to truly replace people," Pearson said. "It might make certain tasks redundant, but it's only going to augment your capabilities." AI will fundamentally change what a workforce does, and perhaps how we think about human capacity and human capital, but it won’t take humans out of the equation.

**ConnectOne Bank is living this out right now**, with revenue rising and plans to hire more people, not fewer. "We could save our frontline folks 50% of their time by taking away the administrative tasks," as CEO Frank Sorrentino told [American Banker.](https://www.americanbanker.com/news/how-connectone-bank-uses-ai-to-save-time-on-admin-work?blaid=8802861) “If we give them 1,000 hours back, does that mean we're going to fire 50% of our producers, or do you think we're going to allow them to go out and be 100% more productive?"

When a technology fundamentally changes how a bank thinks about productivity, like handing people back a thousand hours, the question shifts to **which problems can now be solved with that time.** Opportunities could arise to scale in ways that weren’t addressable or even visible before.

The surest path to a dual workforce runs through the people a bank already has, and how they feel about the change matters as much as how they're trained. As Nicole Caldwell, CMO at nCino, said in a session at nSight 2026, **"I don't think people are scared of change. I think they're afraid of not being taken along for the ride and not being involved."**

For executives, that means bringing along the people on the front lines who feel this change most, keeping their outlook positive, and protecting their institutional knowledge that a bank can't afford to lose. Handled with that kind of care, a workforce grows into the dual future rather than bracing against it.

## AI Is Only as Good as Your Data

Another tension is data. In the benchmark, nearly nine in 10 executives (87%) say they're confident in their ability to access good, quality data, and more than a third are extremely confident. Yet more than half (52%) also report data trapped in silos, along with problems in integrity, consistency, and access. That’s a real disconnect. When confidence outruns the actual state of the data, AI inherits every silo and gap beneath it, and the returns a bank expects fall short.

Confidence outrunning reality is one of the most common blind spots in any organization, because clean-looking dashboards can sit on top of messy data, and problems rarely announce themselves until something built on them underperforms. That's the catch with AI. It's only as good as the data beneath it, so when that data is fragmented or inconsistent, the results are too.

"AI cannot fix underlying structured data,” Pearson explained. “In fact, relying on AI to compensate for gaps, like missing data or siloes, can accelerate data decay. If you have problems with the data that AI consumes, it can translate into poor outcomes.** Every executive should be thinking about their data strategy.**"

Data is one input a bank can fully control. If the foundation is fixed, it lifts the return on everything built on top of it.

## Don't Forget the Fundamentals

A shift this big can make it feel like the old rules no longer apply, but they do. Banking has come through changes like this before, from the arrival of the internet to the move online, and the discipline that carried it through then applies to AI just as well.

That discipline is straightforward.

- Draw a line in the sand and baseline your operations before you change them: how long a loan takes, how many you process, how fast deposits grow.
- Name your path to return, whether that's efficiency, lower overhead, or reduced risk and new revenue.
- Set expectations and a break-even point, then measure against them at steady intervals.

It takes patience, but the opportunity on the other side is worth it, and its full size depends on **treating people and data as seriously as the technology itself.** The return doesn't arrive on a set schedule, and a bank has to trust that the careful work across all three adds up.

Sean Desmond, our CEO, [framed the larger arc.](https://www.americanbanker.com/news/how-connectone-bank-uses-ai-to-save-time-on-admin-work) **"On the other side of every major technology inflection point there has been growth."** Banks are early in this one, and being early is not the same as being behind.

_Find the complete data behind these findings, plus the conversations and solutions shaping the dual workforce at nCino, on the _[_nCino AI in Banking Benchmark_](https://www.ncino.com/ai-in-banking-benchmark)_ page._

---

[View sitemap](https://www.ncino.com/api/markdown/sitemap)