The Shock Absorber

Prefer to listen in your podcast app? Apple Podcasts · Spotify

What is the system trying to teach us?

That's the question behind every Field Note. This week's signal arrived at a bank front desk, after a forty-minute drive.

My husband and I had made an appointment at a bank. We booked it through the bank's app. We had the confirmation. We arrived at the address the app gave us, at the time it told us to be there.

The person at the front desk had no record of us.

He was warm, friendly, and immediately helpful. He gave us exactly what we say we want from someone in a customer-facing role: don't worry, sit down, I'll help you.

It was good customer service.

And it almost made the evidence disappear.

 My husband offered to show him the confirmation on his phone, not to prove that we were right—he kept saying, "No, I want to show you so you can fix your system."

He wanted to show him this evidence that could help him improve the system was right there, sitting in his hand. He wanted to fix their system before he wanted help for what we went there for. (yes, I'm an Ops Nerd married to an Ops Nerd).

At first, the employee didn't need to see it. From his perspective, the problem was already being handled. Two customers were standing in front of him. He was willing to help them. Why spend more time investigating?

Except he couldn't help us. We were there for a conversation that was outside his area of training.

So he looked further. He found the person we were supposed to meet with, called him, and discovered that the specialist covered multiple branches and had gone to a different one.

We hadn't gone to the wrong place. The app had sent us to exactly the right address. The app had sent the specialist to the wrong branch.

No catastrophe. No angry scene. We got back in the car and drove over to meet him.

But I couldn't stop thinking about what had almost happened. A failure in the system had reached the front desk, a kind employee had been ready to solve around it, and if that had worked, the organization might never have known there was anything to learn. For the record, they still might not. We were hand-delivered two workarounds before our meeting was over, so we could bypass the system next time.

But the strange thing about good service is that sometimes the better people are at recovering from a failure, the easier it is for the failure itself to disappear.

I've hired versions of that job. I've managed versions of that job. I’ve had that job. And in healthcare, I've watched organizations struggle over and over to recruit, train, and retain the people who sit where the public meets the operating system.

We describe front-desk jobs as customer-service roles. We look for someone warm, organized, and calm under pressure. Someone who can make a patient feel welcome while answering the phone, finding an insurance card, fixing a scheduling problem, responding to a clinician, navigating a payer question, and figuring out why the computer suddenly insists the patient standing in front of them doesn't exist.

Then we put that person at the point where nearly every failure upstream eventually arrives.

The schedule doesn't work? Front desk. The insurance information is wrong? Front desk. The patient got the wrong instructions? Front desk. The clinician is running behind? Front desk. The technology isn't behaving? Front desk. The handoff between two departments failed? Eventually, somehow, front desk.

And we ask the person sitting there to absorb all of that friction without letting it change them.

We think we've hired a receptionist. We've actually hired a shock absorber.

That is an almost impossible job.

This is where I think leaders can misread what's happening. We're relieved when the front desk solves the problem. Of course we are. The patient gets what they need. The line keeps moving. The phone gets answered. The day survives.

Most of those recoveries never become visible to leadership. Then another version of the same problem surfaces, and another, or someone's out sick for three hours and everything crumbles, and eventually someone asks: why does this keep happening?

Sometimes that frustration lands on the person doing the work. But the employee may be doing exactly what the system has trained and rewarded them to do: resolve the immediate need and move to the next one.

The missing capability isn't kindness. It isn't effort. And it isn't necessarily intelligence. It's the capacity, and the time, to hold two responsibilities at once: solve the problem in front of me, and stay curious about why it appeared.

That second responsibility is systems thinking. Expecting someone to perform it reliably while a patient is waiting, the phone is ringing, a toddler is diligently “choosing” their stickers, and three more problems are forming behind them isn't a training strategy.

It's an infrastructure problem.

This is one of the places I think AI could become genuinely useful in an operating system. Not because the person at the desk should disappear. Because the person at the desk has never really had an operations partner.

Imagine the bank interaction again. Two people arrive with a confirmed appointment that doesn't appear on the local schedule. Instead of requiring the employee to know which systems to check, which branch to call, or which mismatch to suspect, an AI layer could reconcile the confirmation with the calendar, identify the assigned specialist, see that the specialist is somewhere else, and surface the discrepancy.

That gets the customer help faster. But speed isn't the most interesting part.

The more important thing is that the system can preserve the operational breadcrumb the human doesn't have time to chase: an appointment was booked, a confirmation specified this location, the customer arrived here, and the assigned employee arrived somewhere else.

A human trying to keep a line moving may reasonably treat that as a one-off. A system doesn't have to.

It can ask: has this happened before? Was the location presented differently to the employee and the customer? Is there a calendar-sync problem? A workflow problem? A training problem? Is there a pattern across branches? Does someone need to review it?

The human can keep taking care of the person in front of them. The system can keep hold of the unresolved question.

There's a version of this future I don't want: AI becomes the smart layer while the people working inside the organization become less capable of understanding how it works.

I want the opposite. I want AI to build human operational capacity through cues and coaching, not replace it.

We all deserve coaching at work. The front desk is particularly hard to coach because you have to sit alongside the person every minute while they're learning to use the decision trees in the notebook. Once they're familiar, we start to peel away. Once they've built their own understanding and are good enough, we run. Okay, we return for questions and clarification, but no one is happier than the person who was covering the front desk to leave the front desk behind.

I was watching an HBO special about this year's Seahawks training camp. It was a little short on inspiration until we cut to the scene where running back coach Thomas Hammock is coaching his team to establish a physical edge, an identity, a holy vision—literally anything other than last year's rushing offense ranked at the bottom of the NFL. It's noteworthy for the 17 F-bombs in 1 minute, 39 seconds, for sure. But this is the coach we want in our heads, telling us to think, to develop, to grow, to embrace, to be curious, to find new ways to solve old problems. Since we can't import Coach Hammock into every clinic, and since most folks don't respond to inspired action calls drenched in every phrase our mothers banned from our homes, how can we prompt this crucial position into systems thinking without swamping their task list?

Today's AI can easily resolve the mismatch. But I want it to help the employee learn to recognize one. Not with another annual training module, and not by asking someone to stop mid-chaos and perform a root-cause analysis. Inside the work.

At the right moment, the system might come back with something simple: I traced that appointment problem, want to see where the handoff broke? Or: this is the third time this month a customer confirmation and employee location haven't matched, here's what they have in common.

That is more than automation. It's coaching embedded in the workflow. Over time, the employee starts learning the questions the system is asking: what was supposed to happen? What actually happened? Where did those paths diverge? Is this an exception or a pattern? Who else needs to know? And then a weekly report the front desk gives the admin team, so they own the expertise they've been carrying all along.

Systems thinking becomes less of a personality trait we hope someone arrives with, and more of a capability the organization deliberately builds.

In healthcare, front-desk turnover and frustration are often treated as workforce problems. We search again for the right person: warmer, faster, more organized, more resilient, better at multitasking.

Sometimes we find that person. Then we place them inside a job that requires constant invisible repair.

Even an extraordinary employee will struggle in a role that asks them to absorb unlimited operational friction while staying cheerful enough that no one else has to feel it.

So maybe the training question isn't only: how do we teach someone to provide excellent customer service?

Maybe it's also: how do we help the people closest to the friction learn to read the system producing it, one incident at a time?

That doesn't mean making the front desk responsible for repairing the enterprise. It means recognizing that they're standing at one of the richest observation points in the organization, and giving them tools that turn lived friction into operational intelligence, without asking them to carry the analysis alone.

I keep returning to the young man at the bank because I don't think he did anything wrong. I think he did something right. He saw two people with a problem and immediately tried to take care of them.

I don't want to train that instinct out of anyone. I want to add another one beside it: help the person, and hold on to enough curiosity that the organization can learn from what just happened.

Those aren't competing goals. We can provide excellent customer service and investigate the operational question. We can move the line and preserve the evidence. We can solve today's problem without asking the person at the front desk to carry tomorrow's solution alone.

That's the opportunity I see for AI. Not a replacement for the person sitting closest to the customer, although it can and should manage a lot of tasks. A thinking partner for one of the most difficult, under-supported jobs in the organization.

One that remembers what the human doesn't have time to remember. Traces what the human doesn't have time to trace. Sees patterns no individual employee could hold in their head alone. And, if we design it well, helps the human get better at seeing the system too.

Because the person standing closest to the friction shouldn't have to absorb all of it.

They should be among the people best equipped to f*%$@ing help us understand where it came from. Go Hawks.

Keep looking for the signal.

Seen this pattern where you work? Hit reply and tell me. That's the harvest.

Want a bit more? Listen on Apple Podcasts or Spotify for One More Signal about this topic.