In mid-October, I spent two and a half days in Cleveland at MAICON, the Marketing AI Conference put on by Paul Roetzer and the SmarterX team. I sat in on more than a dozen sessions, recorded the ones I could, and came home with a notebook full of ideas.
During his keynote, New York Times tech columnist Kevin Roose shared a tongue-in-cheek list of symptoms for what he calls “feeling the AGI.” They include not being able to think about anything other than AI, dizziness and disorientation, impatience with human inefficiency, and trouble with long-term planning. I’ll admit to a couple of those by the end of Thursday.
As many of you already know, I spent 15 years as a community banker before I started helping banks with their digital strategies - which has evolved to include AI over the past few years. So I sat through every session with one question in mind: what does this mean for a community bank? MAICON is “technically” a marketing AI conference (MAICON literally stands for Marketing AI Conference), not a banking one, so a big part of my job is what I like to call “bankifying” the content: taking what works for marketers and tech companies and translating it into something that makes sense for a bank. Here’s what I took away, and where I think it’s all headed for our industry.
The technology is moving faster than most of us feel it
At the Tuesday half-day workshop, I sat next to a banker friend and couldn’t stop talking about someone I wanted her to meet. Angie Carel from Fort Wayne, Indiana, whom I met at my very first MAICON workshop during my first year at the conference. We became quick friends, and back in June, she delivered a keynote and ran an AI agent workshop for our WSI Global Convention in Mexico.
I told my friend all about Piper, the personal agent Angie had built with Claude and a few other tools. Piper reviewed her calendar, handled her email intake, and worked like a chief of staff whenever someone reached out about a consulting or speaking engagement.
When the three of us finally got together, Angie surprised me that she’d stopped using Piper about a month earlier. She runs everything on Grok Bot now, and she loves it. I’ll admit I was a little taken aback. In June, Piper was the amazing process everyone wanted to copy. By mid-October, “she” had been replaced.
And the AI agent field was already starting to get crowded. Meta’s Muse and OpenAI’s Dots were out before MAICON, and Angie had been experimenting with those too. Then, about a week after the conference, Anthropic launched Claude Duet, its own proactive agent. Knowing what a big Claude fan Angie is, I wouldn’t be surprised if she’s switched a couple of months again from now.
One of the many amazing MAICON speakers, Isabelle Macedo from Google Cloud, put words to what I was feeling. When the speed of AI capabilities doesn’t match the speed of progress inside an organization, it’s hard to build habits. She also described AI as more of an environment than a tool. That’s the lesson for banks: build your strategy, your guardrails, and your philosophy around what you’re trying to accomplish, not around whichever tool happens to be underneath it this quarter.
Roose made the case that by the definitions researchers were using back in 2017, we’ve already reached what they would have called artificial general intelligence. Roetzer went further in his opening keynote, predicting that within the next 12 months AI will handle tasks lasting five hours or more at the level of an A-player on your team.
So why doesn’t it feel that way inside most banks? Roose had a good answer. The capabilities are uneven; most people still use AI as a fancier version of Google, and the gap between what the best models can do and what most people have access to keeps growing. If your only experience with AI is a free chatbot or the feature your core provider added last year, you’re seeing a small slice of what’s possible.
Adopting AI and adapting to it are two different things
Mike Kaput, SmarterX’s Chief Content Officer, shared research showing 93% of marketers now use AI, up from 87% a year earlier. Most of them haven’t changed how they work, though. They’re doing the same things a little faster.
Adopting vs. adapting… what’s the difference? Good question. Adopting AI means buying the tools and using them to speed up the work you already do: drafting the same emails, the same reports, the same marketing pieces, just quicker.
Adapting to AI means rethinking the work itself: changing processes, roles, and decisions because of what AI now makes possible. Adoption gets you a little more efficiency. Adaptation is where the real gains happen, because as Kaput put it, acceleration doesn’t help much unless you’re headed in the right direction.
Dan Slagen from Zapier said something similar: faster creation and better prompts are nice, but the results aren’t showing up where it counts.
That lines up with what I see in my AI advisory work with banks. Buying licenses is the easy part. Getting results from AI is a change management and leadership challenge far more than a technology one, and the organizations that shared their stories at MAICON proved it.
Good Karma Sports, a media company with more than 500 employees, has over 65% of its people using AI every day, including non-technical staff. Their approach was refreshingly simple: set the ground rules up front, give people permission to play and permission to fail, and practice what they called “urgent patience.” Holly Chen walked through how her marketing team at Samsara went from 3% adoption to over 90% in about a year, with non-technical marketers building most of their 100+ agents. One of her ideas I’d love to see in every bank is the “AI Power 30,” a 30-minute, hands-on session once a month with the entire team.
And Sandy Carter shared a stat every bank CEO should hear: companies are 5.2 times more successful with AI when the CEO uses it personally and shares the ups and downs along the way. You can’t hand this one off to IT… let that sink in.
Roetzer made a related point that I haven’t stopped thinking about. In his keynote, he shared a post from Shopify CEO Tobi Lütke from May 9:
"The risk is not that AI does the work. The risk is that AI does the work and we never learn from it. If every interaction with an agent happens in a private window, the only person who learns anything is the person at the keyboard."
The idea is simple. There’s no shame in letting AI do the work, but if you keep how you did it to yourself, you’re the only one who benefits. Roetzer’s takeaway was that people need access to the chain of thought, from both the AI and leadership.
That’s why I encourage every bank to set up an AI council, a working group, or at least regularly scheduled conversations about AI. And those conversations can’t stop at the hours saved or the dollars cut. The real value is in hearing how someone came up with an idea, why they chose a certain path, and what they tried along the way. That’s how one person’s win becomes the whole bank’s capability.
Human in the lead, not just in the loop
By now, I’m sure you’re well acquainted with the phrase “human in the loop.” A variation of that phrase I heard over and over in Cleveland was “human in the lead.” Brian Piper described it as a sandwich. People set the strategic direction, bring the audience insight, and make the judgment calls. AI handles the research, analysis, and scaling. Then people come back in to review, refine, approve, and own the relationship.
One line from the Copilot workshop has stuck with me ever since: “AI owns the remembering. I own the relationship.” Holly Chen said it another way: you can outsource the work, but you can’t outsource the accountability.
Let me bankify that one for you. It’s how a good credit culture already works. An analyst can spread the financials, but the lender and the loan committee own the decision. It’s also exactly the kind of framework your examiners are going to want to see.
Trust is becoming the most valuable thing you own
Kaput shared that Google organic search clicks are down 42% since AI Overviews launched, and about 35% of web pages now show signs of AI authorship. There’s more content than ever and less attention to go around. His advice was to create only what AI can’t: original data, expert interviews, customer stories, and lived experience.
That number should get every bank marketer’s attention. More and more, AI answers the question before anyone ever reaches your website. A business owner asks which local bank is best for an equipment loan, gets an answer, and may never click through. Being found online used to mean ranking on page one. Now it also means being the source AI trusts enough to cite. Chris Penn even pointed to the new generative AI performance report in Google Search Console as a place to start measuring it. If your bank’s SEO strategy hasn’t accounted for AI search yet, now is the time.
Joe Pulizzi took that a step further with what he calls the Trust Portfolio. In a world flooded with AI-generated content, trusted human voices stand out, and he encourages organizations to build a portfolio of them on purpose. That means the expert who makes complicated topics simple, the storyteller from the front line, and the advocate who reminds people why it all matters. Think about who in your organization fits the expert, storyteller, and advocate role. Engage them with your content and publish their perspectives to make a real impact.
Community banks have a head start here. Your lenders, branch managers, and executives are already trusted voices in their communities. The banks that help those people show up online, and that make their websites easy for AI tools to understand and cite, will be the ones that get recommended when a business owner asks ChatGPT where to bank. That’s a big reason I started The LinkedBanker, and it’s a big part of the work we do at WSI.
Where caution is warranted
Not every session was optimistic, and I appreciated that. Carter’s “Five Deadly Sins of AI Agents” reads like a vendor management checklist for banks:
- The Double Agent: an agent that works for you and for whoever pays more.
- The Skeleton Key: one permission that opens every door.
- The Guest Who Never Leaves: an automation nobody remembers to turn off.
- The Empty Chair: leadership that isn’t using the tools.
- No Witness: nobody verifying whether an agent had permission to act.
Penn reminded everyone that generative AI is fundamentally bad at math, so check the numbers before anything goes in front of a board or an examiner. Karen Hao, author of Empire of AI, gave me one of my favorite ways to explain all this. Artificial intelligence is a category, just like transportation. You can get from point A to point B on a bicycle or on a rocket. Sure, both count as transportation, but they are very different things with very different costs.
As an avid cyclist who probably enjoys pedaling more than most who will read this, I was all in on this one. Her argument was for more bicycles: smaller, purpose-built tools (including open-weight models) instead of always reaching for the biggest system available. For banks, that’s a great frame for responsible AI use. Pick the right-sized tool for the job, and be mindful of what it costs, not just in dollars but in data privacy, energy, and impact on the people and communities we serve.
And this isn’t just a feel-good point. Carter noted that agents now use more tokens than humans do, and agentic AI comes with a variable cost that grows with every task it runs. Chen’s team found usage costs ramping fast as adoption spread, so they moved to a monthly AI budget per person with daily tracking. Penn’s advice was to pair a smart model with low effort to keep costs down.
When every employee is choosing between a bicycle and a rocket dozens of times a day, those choices start showing up on the income statement. Even Roose, who is plainly excited about where this is going, called for real regulation, independent evaluators for the AI labs, and better safety controls. And Andrew Yang pointed out that AI will hit hardest the people who trade their time for money, which describes most of us.
What I believe this means for community banking
After two-and-a-half days of drinking from the AI firehose, it’s hard to narrow down my recommendations. There was so much great information. But we can’t boil the ocean, so allow me to share where I think AI will continue to make an impact and change the face of banking over the next few years.
- The gap between banks that adapt and banks that just adopt is going to widen quickly. Two banks can buy the exact same licenses. The one whose CEO is in the tools every week, and whose people have permission to experiment, will pull ahead on efficiency, speed, and talent.
- Community banks will be able to do things that used to require a big-bank budget. Roetzer talked about 100-person companies doing the work of 1,000-person companies. For a $500 million bank, that doesn’t have to mean cutting staff. It can mean growing without adding headcount and giving your best people bigger jobs.
- We’ll have to rethink how bankers learn. This is the one I keep coming back to. Roetzer warned that the traditional career ladder, where junior people learn by doing the tactical work, is starting to break. In banking, that ladder is how credit analysts become lenders and tellers become branch managers.
If AI does the entry-level work, we need a new way to build judgment in the next generation, and the banks that figure out that new apprenticeship model first will win the talent game.
How customers find you is going to change. More of your prospects will ask an AI tool about you before they ever visit your website or walk into a branch. The banks with real people sharing real expertise online are the ones that will show up in those answers.
- And the relationship becomes even more valuable. I’ve said it at a lot of stages this year, and MAICON only reinforced it: AI should make bankers more human, not less. When the machine handles the mechanical work, we get more time for judgment, presence, and the relationships that have always been community banking’s advantage.
Three things to try this week
Roose closed with some simple homework, and I’d pass three of his ideas along to any banker:
- Pay for a good model. The free versions won’t show you what’s possible. Yes, many reading this (assuming you’re in banking) will have access to Copilot. But consider paying for another model on your own. Use it (without bank data, of course, so you don’t violate policy) for some personal tasks and look for differences.
- Give it a hard task, like a real problem you’re working on, instead of a quick email. Think about a process you do at work that takes you multiple steps, hours, and sources of data. Assuming that you’re using an approved tool that can analyze your customer data, this can be a real eye-opener for you.
- Rerun the prompts that didn’t work for you back in 2023. You’ll be surprised. AI continues to evolve, and what didn’t work just a few months ago now works beyond belief. Go back into your chat history and find an old prompt, run it again, and see for yourself how much better (hopefully) the response is. Just imagine what will be possible six months or a year from now!
If you want a baseline, take the AI Score assessment at score.smarterx.ai to see where you and your organization stand. This is a new tool released at the event, along with Paul’s new book, The AI Transformation Blueprint, which we all got copies of, by the way. And if you can get to Cleveland, MAICON 2027 is October 5-7. I’d love to see more bankers in the room.
And if all this seems a bit overwhelming (like maybe your AI Score is not where you’d like it to be), I understand. The good news is you’re just an email or phone call away from someone who can help. Our team works closely with bankers who want to better understand and take their AI strategy to the next level. Get in touch today, and let’s start the conversation.
Embrace digital. Stay human.