If you're a bank executive, asking about return on investment isn't just reasonable. It's your job. Any project that's going to consume time, money, and your people's attention had better produce results, and you have every right to ask any partner you're considering to make the case. So when a bank leader looks at an AI engagement and says, "How do I justify those costs to my board?" - I get it. That's good stewardship.

Here's the problem. The ROI of AI doesn't work like the ROI math you're used to.

For example, when you evaluate digital advertising, the chain is clean. Ad spend generates visits, visits generate applications, applications become loans, and you know what an average mortgage relationship is worth (well, hopefully you do - but that’s a discussion for another day). Apples to apples. You can model it before you spend a dime.

AI is not that animal. The return depends on things a spreadsheet can see up front: your culture, your people's willingness and readiness to implement (and track), whether anyone inside the building actually champions the work and is accountable. Which means anyone who hands you a confident ROI number before understanding your bank - and before your bank understands the technology - is giving you somebody else's results and hoping they transfer. Treat those numbers as suspect.

So in this article, I want to tackle the ROI question the way I wish every executive would hear it: where to expect returns, when to expect them, what has the potential to kill them, and how to build the habit of measuring them for yourself.

Where (and When) to Expect ROI

Let's start with the version your board will recognize - returns by department.

That table is useful. It's also incomplete, because the biggest returns I've seen don't map neatly to an org chart.

Examples of ROI Beyond the Org Chart: Time, Capacity, Capability, and Innovation

One of our banks wanted to truly understand its market. The marketing lead used AI to overlay the bank's own deposit data against census-tract demographics - age stratification across every county in the footprint, penetration by boomers, Gen X, millennials, Gen Z. 

Here's the ROI riddle: what did that analysis save? Nothing. No amount of Excel shenanigans and pivot tables was ever going to produce direct ROI. It's not a faster version of an old task. It's a capability the bank never had before. And it changed where they spend their marketing dollars. How do you put "we stopped advertising into a market we'd already saturated" into an up-front ROI model? You can't. But it's real money.

My own example. I turned my inbox over to an AI-powered dashboard that surfaces what actually matters each morning. Within that surfaced list was a note from a bank that wanted to talk about a small website project. That's a $40,000 opportunity I might have simply scrolled past. What was the ROI on the dashboard? Before that email: "saves me time." After: a number with four zeroes in it.

Lastly, one of our banks has a hotel as a commercial borrower that became delinquent and unresponsive to their outreach efforts. Concerned about the position of their collateral, the banker used AI to mine that property's online guest reviews and spotted recurring complaints about certain rooms - early signals of collateral degradation. That insight became a proactive conversation with the borrower before it became a problem loan. Show me the line in a strategic plan where that was supposed to appear.

And here's the pattern: time saved rarely turns into payroll savings, and that's fine - that's not the point. The point is what fills the hour. The project that's been on the back burner for six months. The relationship call that never got made. You have to count the unlock, not just the hour.

The Barriers that Kill ROI (Nobody Puts These in the Pitch Deck)

So if the returns are real - and they are - why do some banks see them while others end up six months into an AI project wondering what they paid for? Here are the barriers to achieving ROI in a bank.

1. No champion. If nobody inside the bank owns AI adoption, ROI will be sluggish or never arrive because it gets abandoned. I've watched engagements where we lead every meeting, drive every deep dive - and the staff shows up unprepared and doesn't touch the tools between sessions. We can build you a fast car. But if nobody at the bank will drive it, it never leaves the driveway - and that's not an ROI problem with the car.

2. Wins that never get communicated. Team members are trying to move fast and get through their mountain of tasks every day. Within their diligence and speed, they forget to take the time to share their accomplishments with AI or teach workflows with team members. A data analyst at one of our banks hopped on a check-in a few weeks after training. He was beaming with pride because the workflows he'd built were saving him two hours. Per day. For three weeks running. Then I asked: "Does anyone else in the organization know this?" His face dropped. Whoops. He forgot to share what he built and the rewards he was reaping. Thirty-plus hours of documented ROI, invisible to the people who approve the budget - and unavailable to colleagues who'd want some of that for themselves.

3. Normalization. Give it a year and the miracles become staples. One bank's AI champion goes around the room every week asking each person how they're using AI, and the answer is often "oh, not much" - followed by "well, we did automate the whole audit-document process, I guess that saves four hours." Two years ago that was rocket science. Psychologists call this hedonic adaptation: we recalibrate to the new normal and stop counting the win. Your ROI didn't disappear. Your awareness (and appreciation) of it did. If you haven’t baked in a process to communicate actions taken with AI, your ROI will be completely overlooked. 

4. Staff engagement you can't fake. Culture, curiosity, willingness to change - the intangibles every up-front ROI promise assumes.

How to Keep Tabs on ROI: A Methodology

You don't need software for this. You need a habit, at two levels:

The team ritual. Steal the client example I shared above: a standing weekly agenda item where every person answers "what did AI do for you this week?" out loud. It surfaces wins, spreads them, and fights the normalization problem in one move.

The individual habit (with AI doing the tracking). Asking busy people to log their own time savings is a new job nobody wants. So don't. Let the AI keep the ledger. If your team works in a tool like Copilot or Claude, have each person run a Friday prompt like this:

"Review our conversations and my activity from this week. List: (1) the tasks I completed with your help, (2) your best estimate of hours saved versus doing them the old way, (3) anything we produced that didn't exist before, and (4) one win worth sharing with my team. Keep it to five bullets."

Five minutes, every Friday, and suddenly the bank has a running, bottom-up record of exactly the numbers the board keeps asking for. (For Microsoft shops: Copilot's admin analytics and Viva Insights add an org-level layer - usage, adoption, engagement - but tool dashboards measure activity. The methodology above measures value.)

Think Like an R&D Department

One more reframe. Apple and Google pour billions into R&D with no guaranteed return on any given project - and no board on earth tells them to stop, because that spending produced the iPhone and the search engine. Nobody calculates the ROI of R&D one experiment at a time; they judge the portfolio, over time, against the strategic bet that innovation compounds. Early AI work at your bank is closer to R&D than to an ad buy. Fund it like a bet on capability, measure it like a portfolio, and hold it accountable over quarters, not weeks.

The return is real. I've watched it show up as faster loans, smarter markets, protected collateral, and a $40,000 email that didn't get missed. But it arrives on the other side of the work - clear strategy, trained and curious people, wins that get counted - not in front of it.

If you haven't read the companion piece on tying AI to your bank's strategy - start there, because strategy is the first link in this chain.

And if you'd like help clarifying what ROI can look like for your specific bank - and building the organizational habits that make it visible - let's talk. That's exactly the work we do at CommunityBanking.ai. Just get in touch and let's start a conversation.