55 vs 2
It took me 55 minutes to get to a human.
This was last weekend. I noticed a fraudulent charge on my credit card and did what the card told me to do: call customer service. Except, and this is true, I did not reach customer service for almost two hours. Instead, I got stuck in a long, meandering conversation with an AI bot that could not pronounce my name correctly and, more importantly, could not solve my problem.
Eventually, after saying some version of “I want to talk to a human” enough times, I got connected to Debbie. It took her all of 2 minutes to block the charge and send me a new card.
If you take nothing else from this piece, take those two numbers: 55 minutes vs 2 minutes.
That gap says more about the current state of AI in financial services than most product demos ever will.
Where AI Conversations Go Wrong
If you are not running one of the biggest banks in the country or building a fintech designed to remove people from the equation, chances are you still say some version of the same thing: relationships matter. That is what makes so many AI conversations in financial services feel off.
A surprising number of AI deployments begin with the assumption that the goal is to automate the relationship away. The logic is familiar. Reduce calls. Deflect service volume. Replace interaction with automation. Push more activity into channels that require less human involvement. On paper, it sounds efficient. In practice, it often misses the point.
At its worst, this looks like bad automation pretending to be good service. Voice bots with fake warmth. Scripted empathy. Interactions that sound personal while doing everything possible to avoid an actual person. For institutions whose value proposition rests on trust, reassurance, and human judgment, that is not just clumsy. It is strategically backward.
This matters most for community and relationship-led institutions because their value has never been purely transactional. People do not choose them just for a rate sheet. They choose them because they believe there is a human being on the other side who will actually help when something is confusing, stressful, or important.
That does not mean institutions should avoid AI. It means they should be much clearer about where it belongs.
Where AI Actually Helps
The best uses of AI today are usually not the flashy ones. They sit behind the scenes or beside the employee. They help staff find the right policy faster. They summarize calls so people do not spend time on repetitive admin work. They route workflows more intelligently. They flag fraud risk earlier. They surface opportunities or concerns that would otherwise get missed. They prepare employees to have better conversations, not fewer conversations. That is a much better frame for leaders to work from.
The goal is not fewer human interactions. The goal is fewer human minutes wasted on the wrong things.
Routine tasks should move digital when they can. Simple informational requests should become easier to resolve. Internal work that adds no value to the customer should be compressed aggressively. But the time that gets created should not just disappear into a productivity dashboard. It should be reinvested into the moments where people actually want a person: reassurance, edge cases, financial guidance, judgment calls, and complex problem solving.
What Leaders Should Measure
One credit union executive put it to us plainly: “91% of our member transactions are already digital/self-service, yet the contact center still handles 60,000 monthly calls.” The point was not that digital has failed. It was that even in a highly digital environment, people still reach for a human when the moment matters.
That is why too many leaders are still measuring the wrong things. If the only success metric is call deflection, then almost any automation can look smart for a quarter. The better metrics are harder and more important: resolution quality, trust, escalation quality, employee leverage, fraud prevention, and whether staff now have more time for higher-value conversations.
That is where leaders should focus. Not on whether AI can mimic a human well enough to pass. On whether it allows employees to do more of the work that actually requires one.
Human Banking
The right AI deployment gives people more room to be human. The wrong one asks them to disappear. That is the real distinction.
Good AI should help your team become more useful, more prepared, and more responsive. It should shorten the path to resolution. It should remove friction behind the scenes so the person on the other end can do what Debbie did for me in 2 minutes.
That is the standard.
Not whether the bot sounded friendly. Not whether the dashboard shows fewer inbound calls. Not whether the automation looked impressive in a demo. Whether the institution used technology to get a customer to the right outcome faster, with more confidence and less frustration. The institutions that win will not be the ones that automate human service away. They will be the ones that use AI to make human service more valuable.
Because in the moments that actually matter, nobody remembers the 55 minutes fondly. They remember Debbie.
Stats That Matter
- 77% of people have gone digital for routine banking. Mobile and online now dominate how people manage accounts. Branches and phones are no longer the default for simple tasks. source
- Only 18% of people trust AI alone for financial recommendations. Consumers are open to AI assistance, but not to removing human accountability. source
News that Matters
- "The Deflection Trap": A recent Financial Brand report found that while 91% of bank boards have approved Gen-AI programs, the most successful institutions are shifting metrics from 'Call Deflection' to 'Resolution Quality.' source
- "Humanity as a Strategy": On a recent episode of The Outlier Podcast, industry leaders argued that the "grey areas" of banking—fraud, life events, and complex lending—are where credit unions will win by using AI to "shorten the path to a human," not lengthen it. source
- "Efficiency vs. Empathy": CU 2.0 warns that credit unions focusing solely on automation risk losing their "relationship premium," suggesting that AI’s best role is in the back office, making the front office more "human." source
