Industry insights
AI will not shrink your bank. It will multiply what your people can do.
Michael Abbott's July piece in Forbes, "Why AI Could Unlock Up To 40% More Capacity At Banks," makes an argument we have been making to credit unions and community banks for three years, with a much bigger dataset behind it. The headline number comes from Accenture analysis: AI could unlock roughly 40% more capacity across financial services.
Capacity. Not headcount reduction. That distinction is the whole article, and it is worth sitting with.
Every previous wave was supposed to shrink the bank. None of them did.
Abbott walks through the history. Tabulating machines rewired clerical work in the 1930s. Mainframes automated core processing in the 1960s and 70s. Online and mobile banking moved whole categories of teller transactions onto the phone in the customer's pocket. Each wave arrived with the same assumption: this technology exists to replace people and cut operating cost.
And yet FDIC data shows U.S. banking employment has nearly doubled since the 1960s, from roughly one million to just under two million by the end of 2024. The work changed. The people stayed. There were simply more things worth doing once the drudgery got cheaper.
We think AI follows the same pattern, for the same reason: the constraint at most institutions was never the number of employees. It was how much any one employee could hold in their head.
The number that should stop you: 75%
Abbott cites relationship managers spending up to 75% of their time on administrative work — preparation, research, summaries, follow-ups, compliance steps. Three quarters of the day belonging to the job around the job.
Ask a branch manager at a community bank how their newest personal banker spends an afternoon and you will hear a smaller version of the same story. Not the conversation with the member in front of them. The hunt for which version of the disclosure applies, which exception needs a second signature, whether the Reg CC hold rule changed last quarter.
The Forbes framing is that if you hand each adviser the equivalent of a support team, you do not get fewer advisers. You get advisers who can actually serve the book they already have. Abbott points to JPMorganChase, where the bank reports AI agents let each private banker handle roughly 50% more clients.
Where this gets real for a community institution
The multi-agent future Abbott describes is genuinely coming, and a $900 million credit union is not going to build it. But the first and largest slice of that 40% does not require it. It requires answering the question every employee asks a dozen times a day: where is the current, correct answer, and can I trust it enough to say it out loud to a member?
That is precisely what CurrentWave Portal does. Your policies, procedures, product guides, and disclosures become something your team can ask in plain language, and every answer cites the document it came from.
At Marine Credit Union, the search for an answer went from about two minutes of hunting to 4.24 seconds. That started in the call center and spread to all 16 branches. Two minutes does not sound like a crisis. Multiply it by every question, every employee, every day, and you have found real capacity inside a staff you already pay — no requisition, no recruiting cycle, no ramp.
The part nobody talks about: your newest employee
The most quietly important passage in Abbott's article is about talent development. If AI performs much of the entry-level work, learning by doing declines, and training has to shift from technical skill to judgment. He suggests banks may move toward simulation-based development, the way aviation does.
Our experience points somewhere slightly different, and more immediately useful. A new hire with a sourced assistant does not stop learning. They learn faster, because every answer arrives attached to the document it came from. They are not memorizing a binder. They are reading the actual procedure, in context, at the exact moment it matters, twenty times a day.
The traditional path was: shadow someone experienced for six weeks, then interrupt them for six months. That path costs the institution twice — once for the new employee's slow ramp, and once for the tenured employee's constant interruption. A new personal banker in week two, with the current answer and its source in front of them, sounds like a banker in year two. And the person who used to be interrupted gets their afternoon back.
This is the version of "more capacity" that a community institution can act on this quarter. Not replacing your people. Making your least experienced employee substantially more capable, and freeing your most experienced one to do the work only they can do.
FTE is the wrong metric. What should you measure instead?
Abbott argues the question is no longer how many people a bank employs but how much capacity it creates. We would make that concrete for an institution your size. Four things worth tracking:
- Time to a trustworthy answer. Not "did they find something," but how long until a frontline employee has an answer they will say out loud to a customer.
- Interruption rate. How often does a frontline question escalate to a tenured colleague, a supervisor, or compliance? That number falling is capacity appearing.
- Time to productive for a new hire. The week at which a new employee handles a normal queue unaided.
- Answer consistency across locations. Ask the same question at three branches. If you get three answers, you have a risk problem wearing a training problem's clothes.
None of these show up on an FTE line. All of them show up in what your members experience.
Humans in the lead
Abbott's closing point is the one we would underline: the banks that move fastest will be the ones that keep humans in the lead while using AI to expand capacity, improve decisions, and refocus talent on the work where judgment, trust, and relationships matter.
We would only add that this is not a distant strategic posture. It is a decision about where the answer lives. If it lives in a shared drive, an email thread, and the head of the person who has been there nineteen years, your capacity is capped at whatever that person can personally handle in a day. If it lives somewhere every employee can ask, your capacity is a different number entirely.
The 40% figure is a projection. The 4.24 seconds is not.
What would 40% more capacity look like at your institution?
Tell us how your team works and we will walk you through what they would be asking, and what they would get back. Most institutions are live in about one business day.
References
- Abbott, M. (2026, July 23). Why AI could unlock up to 40% more capacity at banks. Forbes. forbes.com