Let me start with a question.
When was the last time a company failed you… and you ended up loving them more for it?
Take a second. Because if you’ve got an answer—and most of us do—I promise you it didn’t happen by accident. It happened because somewhere in the wreckage of a botched order, or a missed delivery, or a chatbot that looped you in circles for twenty minutes, a human being stepped in and did something so right that the failure itself became the setup for the save.
You didn’t just forgive them. You started telling people about them.
Now, psychologists like me have a name for this strange little turn, and so do service researchers. We call it the service recovery paradox. And the finding is this: a customer whose problem gets resolved brilliantly can end up more loyal than a customer who never had a problem at all.
Not despite the failure. Because of how it was handled.
I’ve spent nearly four decades consulting inside organizations people love— companies like Starbucks, The Ritz-Carlton Hotel Company, Mercedes-Benz, Zappos, and Amazon One Medical. And here’s a truth that survives every era and every new piece of technology: nobody is loyal to a company that has never been tested.
Loyalty doesn’t get forged when everything goes right. It gets forged the moment something goes wrong—and the customer finds out who you really are. I call that pivot the rescue moment. And in the AI era, I think it’s becoming one of the most important moments your brand will deliver.
Let me name the uncomfortable thing that leaders frequently tiptoe around.
As AI takes over more of your customer experience, failures don’t go away. In some ways, they multiply. In others, they just change shape.
The bot misreads the question. The automated flow has no path for the customer’s actual situation. The self-service portal solves nine problems beautifully… and strands the tenth customer in a hall of mirrors with no door marked human.
Here’s the thing about automation. It is extraordinary at handling the expected. And it is brittle—sometimes cruelly so—at handling the exception. And the exception is exactly where emotion runs highest and loyalty is most on the line.
So the question is not “how do we eliminate failure?” You won’t. The real question—the honest one—is: when our system breaks, and it will, what happens in the ninety seconds after? Because that’s the moment your customer learns whether they’re dealing with a brand that cares, or a machine that processes.
Let me give you the data, because the case for great recovery is not sentimental. It’s financial, and it goes back decades.
Start with the foundational work of John Goodman and a research firm called TARP. They ran the White House–commissioned studies that basically set the modern standard for how companies handle complaints. And Goodman’s research surfaced a finding that should be on every service leader’s wall: a customer who complains and gets a satisfying resolution is roughly thirty percent more loyal than a customer who never had a problem in the first place.
Sit with that for a second. A well-handled failure doesn’t just get you back to neutral. It can leave the customer more attached than smooth sailing ever would.
And Goodman found something else just as important—the silence problem. The overwhelming majority of wronged customers never complain at all. They just… leave. And they tell other people on the way out. Which means the customer who does bother to complain is not your enemy. They’re handing you a rare and expensive gift: a chance to show, under pressure, exactly who you are.
There’s more. Qualtrics research found that nearly half of consumers—forty-eight percent—said they liked a retailer more after an issue got resolved quickly and well. That’s the recovery paradox, caught in the wild. And on the loyalty side, Salesforce research has reported that around seventy-eight percent of customers who get a satisfying resolution stay loyal. That turns straight into repeat business and word of mouth.
So the recovery, done right, is one of the highest-return moments in the entire customer journey. And most companies treat it like a cost center.
Now, why does AI make this the whole ballgame? Three quick shifts.
First: speed of fix is becoming table stakes. AI can resolve routine problems faster than any human—instant refunds, automated reorders. That’s wonderful. But when everyone can do the easy stuff instantly, “fast” stops winning anyone’s heart. What’s left is how the hard failure gets handled.
Second: the failures that reach a human are now the worst ones. As bots absorb the simple complaints, the issues that finally escape to a person are the gnarly, emotional, edge-case ones—after the customer’s already been failed twice. Once by the original problem. Once by the bot that couldn’t help. The emotional stakes of every escalated contact have gone up, not down.
Third—and this one matters most: accountability cannot be automated. Customers don’t just want their problems solved. They want someone to own them. “I’m sorry this happened, this one’s on me, here’s what I’m going to do”—that’s a deeply human act. It requires a self to take responsibility. An AI can apologize in flawless, frictionless prose, and customers can feel the emptiness of it in about a second.
Let me make this concrete with a story that became legend precisely because it broke every efficiency rule in the book.
At Zappos—the online retailer whose whole brand was built on service and one I’ve worked with and written the book The Zappos Experience book about - their longest recorded customer-service call ran ten hours and forty-three minutes.
Ten hours.
Most contact centers would treat that as a catastrophe. Handle time is the metric they live and die by. Zappos celebrated it. Their founder, Tony Hsieh, said plainly that an employee’s job was never to get the customer off the phone fast. It was to make the customer happy—however long that took.
Now, that headline call became two humans connecting over shoes and life. But sit with what the policy underneath it reveals, because that’s the lesson. Zappos deliberately refused to let efficiency metrics govern the human moments that build loyalty. They understood that when a customer’s on the line—especially a frustrated one—the worst possible message you can send is “we’re trying to get rid of you quickly.”
The brands that win recovery do the opposite of what the dashboard begs them to do. They slow down. They stay present. They treat the problem as a relationship, not a ticket to close. And that’s exactly the instinct AI threatens to erode—because automation is engineered to minimize handle time, and recovery sometimes requires the courage to maximize human attention at the precise moment a spreadsheet says minimize it.
So how do you make recovery a strength instead of a scramble? Three moves.
One: pre-decide the escape hatch. Map your automated journeys, mark every place a customer could get stuck, and build a fast, obvious, low-friction path to a competent human. Not a buried “contact us” link. A real door.
Two: empower the recovery—don’t script it. A frustrated customer can tell the difference between a person reading an apology and a person taking responsibility. Give your frontline the authority to fix things on the spot. Recovery dies in the words “let me check with my manager.”
Three: measure the save, not just the speed. If your dashboard tracks handle time but never asks whether the recovered customer felt genuinely cared for, you’re optimizing for the wrong thing.
So here’s the reframe I’ll leave you with.
Stop treating service failure as a defect to be minimized. Start treating recovery as a moment to be won. In a market where AI is rapidly turning smooth service into a commodity, the broken moment—the one your competitors are automating into an even colder dead end—is the richest opportunity you have left to prove you’re human, that you care, and that the person on the other end is more than a ticket.
Failures are inevitable. More so now than ever. Loyalty is not made in their absence. It’s made in the rescue.
So the question I want you to sit with is this: when your system fails a customer tomorrow—and somewhere, it will—is the next thing they touch a deeper, colder dead end? Or a human who turns the worst moment of their week into the reason they’ll never leave you?
If designing that rescue moment into your customer experience is work you need to do, that’s exactly what I help leadership teams achieve. Please reach out to me at josephmichelli dot com slash contact. And let’s have a conversation so service recovery is your most loyal advantage.
Thanks for your time. I’ll see you next week.
When was the last time a company failed you… and you ended up loving them more for it?
Take a second. Because if you’ve got an answer—and most of us do—I promise you it didn’t happen by accident. It happened because somewhere in the wreckage of a botched order, or a missed delivery, or a chatbot that looped you in circles for twenty minutes, a human being stepped in and did something so right that the failure itself became the setup for the save.
You didn’t just forgive them. You started telling people about them.
Now, psychologists like me have a name for this strange little turn, and so do service researchers. We call it the service recovery paradox. And the finding is this: a customer whose problem gets resolved brilliantly can end up more loyal than a customer who never had a problem at all.
Not despite the failure. Because of how it was handled.
I’ve spent nearly four decades consulting inside organizations people love— companies like Starbucks, The Ritz-Carlton Hotel Company, Mercedes-Benz, Zappos, and Amazon One Medical. And here’s a truth that survives every era and every new piece of technology: nobody is loyal to a company that has never been tested.
Loyalty doesn’t get forged when everything goes right. It gets forged the moment something goes wrong—and the customer finds out who you really are. I call that pivot the rescue moment. And in the AI era, I think it’s becoming one of the most important moments your brand will deliver.
Let me name the uncomfortable thing that leaders frequently tiptoe around.
As AI takes over more of your customer experience, failures don’t go away. In some ways, they multiply. In others, they just change shape.
The bot misreads the question. The automated flow has no path for the customer’s actual situation. The self-service portal solves nine problems beautifully… and strands the tenth customer in a hall of mirrors with no door marked human.
Here’s the thing about automation. It is extraordinary at handling the expected. And it is brittle—sometimes cruelly so—at handling the exception. And the exception is exactly where emotion runs highest and loyalty is most on the line.
So the question is not “how do we eliminate failure?” You won’t. The real question—the honest one—is: when our system breaks, and it will, what happens in the ninety seconds after? Because that’s the moment your customer learns whether they’re dealing with a brand that cares, or a machine that processes.
Let me give you the data, because the case for great recovery is not sentimental. It’s financial, and it goes back decades.
Start with the foundational work of John Goodman and a research firm called TARP. They ran the White House–commissioned studies that basically set the modern standard for how companies handle complaints. And Goodman’s research surfaced a finding that should be on every service leader’s wall: a customer who complains and gets a satisfying resolution is roughly thirty percent more loyal than a customer who never had a problem in the first place.
Sit with that for a second. A well-handled failure doesn’t just get you back to neutral. It can leave the customer more attached than smooth sailing ever would.
And Goodman found something else just as important—the silence problem. The overwhelming majority of wronged customers never complain at all. They just… leave. And they tell other people on the way out. Which means the customer who does bother to complain is not your enemy. They’re handing you a rare and expensive gift: a chance to show, under pressure, exactly who you are.
There’s more. Qualtrics research found that nearly half of consumers—forty-eight percent—said they liked a retailer more after an issue got resolved quickly and well. That’s the recovery paradox, caught in the wild. And on the loyalty side, Salesforce research has reported that around seventy-eight percent of customers who get a satisfying resolution stay loyal. That turns straight into repeat business and word of mouth.
So the recovery, done right, is one of the highest-return moments in the entire customer journey. And most companies treat it like a cost center.
Now, why does AI make this the whole ballgame? Three quick shifts.
First: speed of fix is becoming table stakes. AI can resolve routine problems faster than any human—instant refunds, automated reorders. That’s wonderful. But when everyone can do the easy stuff instantly, “fast” stops winning anyone’s heart. What’s left is how the hard failure gets handled.
Second: the failures that reach a human are now the worst ones. As bots absorb the simple complaints, the issues that finally escape to a person are the gnarly, emotional, edge-case ones—after the customer’s already been failed twice. Once by the original problem. Once by the bot that couldn’t help. The emotional stakes of every escalated contact have gone up, not down.
Third—and this one matters most: accountability cannot be automated. Customers don’t just want their problems solved. They want someone to own them. “I’m sorry this happened, this one’s on me, here’s what I’m going to do”—that’s a deeply human act. It requires a self to take responsibility. An AI can apologize in flawless, frictionless prose, and customers can feel the emptiness of it in about a second.
Let me make this concrete with a story that became legend precisely because it broke every efficiency rule in the book.
At Zappos—the online retailer whose whole brand was built on service and one I’ve worked with and written the book The Zappos Experience book about - their longest recorded customer-service call ran ten hours and forty-three minutes.
Ten hours.
Most contact centers would treat that as a catastrophe. Handle time is the metric they live and die by. Zappos celebrated it. Their founder, Tony Hsieh, said plainly that an employee’s job was never to get the customer off the phone fast. It was to make the customer happy—however long that took.
Now, that headline call became two humans connecting over shoes and life. But sit with what the policy underneath it reveals, because that’s the lesson. Zappos deliberately refused to let efficiency metrics govern the human moments that build loyalty. They understood that when a customer’s on the line—especially a frustrated one—the worst possible message you can send is “we’re trying to get rid of you quickly.”
The brands that win recovery do the opposite of what the dashboard begs them to do. They slow down. They stay present. They treat the problem as a relationship, not a ticket to close. And that’s exactly the instinct AI threatens to erode—because automation is engineered to minimize handle time, and recovery sometimes requires the courage to maximize human attention at the precise moment a spreadsheet says minimize it.
So how do you make recovery a strength instead of a scramble? Three moves.
One: pre-decide the escape hatch. Map your automated journeys, mark every place a customer could get stuck, and build a fast, obvious, low-friction path to a competent human. Not a buried “contact us” link. A real door.
Two: empower the recovery—don’t script it. A frustrated customer can tell the difference between a person reading an apology and a person taking responsibility. Give your frontline the authority to fix things on the spot. Recovery dies in the words “let me check with my manager.”
Three: measure the save, not just the speed. If your dashboard tracks handle time but never asks whether the recovered customer felt genuinely cared for, you’re optimizing for the wrong thing.
So here’s the reframe I’ll leave you with.
Stop treating service failure as a defect to be minimized. Start treating recovery as a moment to be won. In a market where AI is rapidly turning smooth service into a commodity, the broken moment—the one your competitors are automating into an even colder dead end—is the richest opportunity you have left to prove you’re human, that you care, and that the person on the other end is more than a ticket.
Failures are inevitable. More so now than ever. Loyalty is not made in their absence. It’s made in the rescue.
So the question I want you to sit with is this: when your system fails a customer tomorrow—and somewhere, it will—is the next thing they touch a deeper, colder dead end? Or a human who turns the worst moment of their week into the reason they’ll never leave you?
If designing that rescue moment into your customer experience is work you need to do, that’s exactly what I help leadership teams achieve. Please reach out to me at josephmichelli dot com slash contact. And let’s have a conversation so service recovery is your most loyal advantage.
Thanks for your time. I’ll see you next week.
