People have been predicting the end of community banks for as long as I have been paying attention. Too small to compete, too slow to modernize, about to be swallowed by the big banks or routed around by fintechs. The pressures behind those predictions are real. And yet the obituary keeps turning out to be early. Having spent time in this part of the industry, I have come to see why, and what tends to set the community banks that endure apart.

Instant payments have quietly crossed the line from launch to infrastructure. Most of the country can reach a real-time rail now, which was not true a couple of years ago. But from where I sit, having spent a chunk of my career in payments, the milestone is not the interesting part. The interesting part is the gap between being connected to a rail and actually running meaningful volume across it, and the operational weight that real-time quietly piles onto a bank.

I still read funding flows out of habit, a leftover from years of raising money myself. That instinct has made the AI wave more interesting to watch than anything has been in a long time. What I am seeing is a familiar pattern at a scale I have not seen before. The money is pouring in, and it is pouring into a few places while running right past others.

Plenty happened in bank technology this year, and most of it was noise. Rather than cataloging all of it, I want to name the two shifts that actually changed how I think, the ones that still felt real once the conference hype wore off. They are not the flashiest stories of the year. They are simply the two that mattered.

I have spent enough time building with AI agents on my own projects to be genuinely impressed by them. I have also spent enough time around banking to know that the harder I imagine putting one to work inside a bank, the faster I hit a wall, and the wall is not technical. It is governance. The moment software stops suggesting and starts acting, the frameworks a bank relies on to stay safe begin to strain.

Distance changes how you read your own history. With some years between me and the startup chapter, and a stretch on the other side of the table inside a bank, a lot of what used to feel like friction or bad timing looks different now. Some of the walls I spent energy resenting turned out to be the actual shape of the business. Here is what I would tell the version of me who was building back then.

The thing that finally made AI agents click for me was a small distinction. A chatbot can tell you how to reverse a fee. An agent can go reverse it. One explains the steps. The other takes them. Once I started playing with that on my own projects, the shift stopped feeling incremental and started feeling like a different category of tool.

When ChatGPT first showed up, I treated it like a magic trick. I'd throw prompts at it to see what it could do, marvel for a minute, and move on. Somewhere in the year and a half since, that changed without my really noticing. On my own projects, on my own time, these tools stopped being a novelty and became part of how I actually work. Not a thing I test. A thing I reach for.

For a few years, Banking as a Service looked like the rare arrangement where everyone won. Fintechs got to offer bank accounts without a charter, small banks got fee income and a foot in the digital door, and the people using the apps got slick products. I wanted that arrangement to work. Years ago, when I was building a payments product, a partnership like this was exactly what I was looking for and couldn't get. So when the model started breaking in the most visible way possible, I paid close attention, this time from inside a bank rather than outside trying to get in.

For a while now, operations has struck me as the place where technology could do the most good inside a community bank. Not the headline projects. The day to day work of serving clients and moving loans through the pipeline. The easy assumption is that more automation is the goal in itself. I see it differently. The point of operations technology is to make the bank faster, more responsive, and easier to work with, for clients and for the people serving them. That is a different target than automation for its own sake.