The Briefing · Issue 4 · June 15, 2026 · 6 min read
Buying the Tools Was the Easy Part
You already pay for the AI. Whether anyone actually uses it well is a people problem, and the data on that just got a lot clearer.
Here’s a conversation I have all the time. A bank tells me they’ve “done AI.” They bought the licenses, sent out the announcement, maybe ran a lunch-and-learn. Then they look at the usage numbers a few months later and quietly wonder why almost nothing changed.
I’ve watched this enough times to tell you the uncomfortable truth. Buying the tool is the easy part, and it’s the part everyone spends their money on. The hard part, the part that actually determines whether you get any return, is human. It’s adoption. And adoption doesn’t come in a license.
Someone finally put numbers to it
I’ve believed this from watching bankers, but belief is not data. So I was glad to see Paul Baier of GAI Insights publish a piece called “What 1,720 Wall Street Employees Taught Me About Learning AI.” His team ran 53 AI workshops for eight private equity firms, hedge funds, and asset managers over seven months, training 1,720 employees. Full credit to Paul for the work and the numbers, and the link is below.
Now, that’s a Wall Street crowd, not a community bank crowd, and Paul says as much. But strip away the industry and what he documented is human behavior around a new tool, and that generalizes to your shop more than you’d expect. A few of his findings stopped me cold.
“I use AI daily” tells you almost nothing. Most people who say it are using AI as a fancier search engine. Very few have ever set up the custom instructions, projects, or reusable setups that separate a party trick from real leverage. Access is not fluency, and your daily-active numbers are hiding this.
A tiny group carries everything. Paul found that 3 to 5 percent of employees, the power users, drive more than 70 percent of the usage. That held true even at firms spending hundreds of thousands of dollars on training and hackathons. Money does not solve a change-management problem.
Generic training produces little value. A tour of features doesn’t stick. What works is training tied to a person’s actual job and your bank’s actual approved tools and policies.
The CEO sets the pace. Paul found that within the same industry, the gap between firms traces directly back to how much the chief executive personally uses the tools. A CEO who actually works with AI sets the speed, the budget, and the urgency.
A CEO who delegates the topic sends a quiet signal that it's optional, and the organization reads that signal perfectly.
Why this is good news for you
If AI success came down to who bought the best software, the biggest institutions would win every time and you would lose. But it doesn’t. It comes down to culture, habits, and leadership attention, and those are exactly the things a community bank is good at. You already know how to build a culture. You already know your people by name. That is the whole ballgame here, and it’s a game you can win.
But how does this apply to my bank
Here’s what actually moves adoption, drawn from Paul’s findings and from what I see working inside banks.
Go first, visibly
If you're the CEO or a senior leader, your personal use is the single biggest lever you have. You don't have to be an expert. You have to be seen using it, talking about it, and asking your team what they've figured out. Nothing in a policy memo comes close.
Name your champions and reward sharing
Find the 3 to 5 percent who are already into it and give them a real role. Then make sharing a tip something that gets you noticed in a good way, not something people hoard to look smart at review time. Paul is blunt that a sharing culture and a champions network beat every other tactic he measured.
Train on their work, not on features
Skip the generic tool tour. Sit a lender down with a real memo, a marketer with a real campaign, an ops person with a real pile of documents, and train on that. It has to map to the job and to the tools you've actually approved.
Make it ongoing, not one-and-done
Paul's finding that a train-once approach fades within weeks matches everything I see. A short, regular rhythm of coaching beats a big one-time event every time.
Start with the people drowning in documents
Your biggest unlock isn't your engineers, it's the 95 percent who don't code and are buried in email, spreadsheets, and PDFs. Give them one real, useful project and let them feel the time come back.
On that last point, let me give you a concrete first project, because “start using AI” is too vague to act on. One place I focus a lot is financial literacy, the community education work your team gets tapped for every year. It’s ideal AI territory: real output, genuine community value, and almost no risk, since you’re creating educational material, not touching customer data. AI can act as your curriculum designer, content creator, and practice partner, and take that program from a someday item to a done one. I wrote a full how-to on it for The Financial Brand, with ready-made prompts, and it’s linked below as a place to actually begin.
The banks that win with AI over the next two years will not be the ones who bought the most.
They’ll be the ones whose leaders went first, whose culture rewarded sharing, and who trained people on the work in front of them. None of that is on a price sheet. All of it is within your reach.
If you want help building the adoption side, the champions, the coaching rhythm, the first real projects, that’s exactly the work I do with banks. Let’s talk.
Build your AI adoption plan with me →
Worth reading
Ben