Every board deck now seems to have an artificial intelligence slide, and the pressure to show progress is real.
Seventy-two percent of bank executives and directors say their bank has implemented generative AI, and 30% are using agentic AI, according to Bank Director’s 2026 Technology Survey. Getting AI in the door isn’t the hard part.
Yet when an AI pilot stalls, the instinct is to blame the tool. More often, the tool did its job but also shined a light on what was already broken. The gap usually isn’t the technology but the way the institution operates.
Outdated Processes, Fragmented Systems and Siloed Data
AI can speed up a broken process, but it can’t fix one. In the same Bank Director survey, 27% of respondents cite obsolete technology that requires workarounds as a top concern. Addressing that takes modernization: replacing manual steps so the institution runs the way it should.
Then there are fragmented systems and siloed data. The survey also found that 69% of respondents name integrating new technology with existing systems as a top concern, up 18 percentage points in a year. Sixty-one percent keep data in the system that generates it, and 55% use spreadsheets to manage business-line data. Connecting systems and the data within gives AI a complete picture with which to work.
These problems are familiar. The next two are harder to see, and AI is what makes them visible.
Everyone Has an AI Pilot. No One Has the Same Destination.
Picture a bank or credit union where marketing is testing generative AI for campaign copy, operations is automating reporting and compliance is trialing a monitoring tool. Each project makes sense on its own. None is tied to a shared goal. The institution is automating activity, not moving an outcome.
Bank Director’s survey suggests this is common: among banks using generative AI, 80% apply it in operations, 73% in marketing and 66% in compliance. Yet 76% of respondents see technology primarily as an engine for efficiency, while just 24% view it as a tool to generate growth. Adoption is broad, but the ambition is narrow.
The real question isn’t how many AI projects are running. It’s how much value the institution gets from its data and AI. The payoff comes from moving beyond doing things faster to knowing what to do, and who should do it.
Insight Nobody Acts On
Consider a longtime customer whose activity is quietly dropping. An AI model flags it. The alert lands in a dashboard. Marketing assumes lending will call. Lending assumes the branch will. The branch never sees it. The insight was generated. The action never followed.
It’s one thing to uncover a trend in your data. It’s another to have the data tell you what to do about it. That’s the difference between reporting and activation. Activated data doesn’t just flag the at-risk customer. It recommends the next best step, such as a personal call or a tailored offer and puts it in front of a specific person in their workflow: a prompt for the teller or an alert for the loan officer. Then it tracks the outcome.
It Takes Culture, Not Just Code
None of this means a team or leader got it wrong. Silos, unclear priorities and unowned handoffs are usually the natural result of years of growth, new vendors and workarounds that made sense at the time. They reflect how the organization works together, which is why another tool won’t fix them.
Boards and executive teams set the tone by making it safe to surface what AI reveals and treating it as a chance to improve, not as a failure.
Questions Worth Asking
- Where do our most important decisions still depend on spreadsheets or manual processes?
- Which business goal is each AI initiative meant to move, and is that goal shared across departments?
- Which one or two friction points would move the needle most if fixed first?
Fix the Foundation First, Then Apply AI
Modernize operations so the institution runs the way it should. Connect systems so everyone works from the same picture. Activate data so the right insight reaches the right person at the right moment, with shared goals and clear ownership.
No financial institution needs all three in place to start, and each step builds on the last.
Previous