Efficiency Didn't Save Your Customer Relationship. It Killed It.

Something has been quietly shifting.

Not just in technology. In us.

I've spent 25 years in rooms where people make decisions about other people. About how many agents to hire. Which roles to automate. When a customer interaction is "worth" a human versus a machine. And what I'm seeing now is something I haven't seen before.

For the first time, we are genuinely unsure where humans fit.

That's not a technology problem. That's an identity problem. And it's bigger than customer service.

Think about what the last three years have asked of us.

AI can now write. It can reason, summarize, recommend, and respond. It can pass bar exams, draft strategy decks, and hold a phone conversation without flinching. The things we told ourselves made us irreplaceable — critical thinking, communication, problem-solving — are no longer exclusively ours.

So we're all quietly asking the same question, whether we say it out loud or not.

What exactly am I for?

That question lives in the C-suite. It lives on the call center floor. It lives in the agent who wonders if she'll have a job in two years. It lives in the executive who approved the AI budget and is now watching CSAT quietly decline.

We haven't resolved it. We've just gotten better at not talking about it.

Here's what I keep seeing in the operations and companies that I have come across.

Companies move fast on AI. They automate the volume, cut the headcount, celebrate the efficiency metrics, and declare victory. Then, 12 to 18 months later, something starts to feel off. The dashboard still looks clean. But repeat contacts are climbing. Customer complaints are getting angrier. The interactions that used to get resolved are now getting escalated to agents who are stretched thin and inheriting frustrated customers with no context.

The AI handled the transaction. Nobody handled the relationship.

And the customers noticed before leadership did.

This pattern has played out publicly at some of the world's most watched companies. The reversals are now too consistent to ignore.

Klarna replaced roughly 700 customer service workers with AI, watched their efficiency metrics soar, and then spent the next year quietly rebuilding human capacity after customer satisfaction deteriorated on anything complex or emotionally charged. Their CEO admitted the company had gone too far, saying cost had become too dominant an evaluation factor. The result was lower quality.

The Commonwealth Bank of Australia ran the same experiment on a smaller scale. They replaced over 40 customer service staff with an AI voice bot. The system couldn't perform the job as well as humans, call volumes increased rather than decreased, and the bank reversed its decision. They later admitted they did not adequately consider all relevant business considerations when making the call.

IBM went further into their own operations. Their AI-powered internal platform now automates 94% of routine HR inquiries. But the remaining cases, the ones requiring judgment and context, still depend entirely on human employees. IBM's response? They announced plans to triple entry-level hiring for roles that require human judgment, customer interaction, and oversight of AI systems.

Three different companies. Three different industries. One identical lesson.

They all discovered, too late, that their humans weren't just handling interactions. They were holding the relationship together.

Reading the tone of a frustrated customer and adjusting. Knowing when policy said one thing but the right answer was something else. Being the two-year agent who has heard every version of a billing complaint and knows exactly which sentence de-escalates it before it becomes a cancellation.

You can't document that. You can't train a model on it in six months. It lives in people.

And when you remove the people, you don't just lose the labor. You lose the institutional memory, the emotional intelligence, and the judgment that no one thought to measure because it never showed up in a cost-per-call report.

Gartner put it plainly. Don't assume that because AI can access everything your people know, it can do everything people do. The companies reversing course replaced jobs that required human judgment and got information retrieval instead.

So what does it mean to be human in a world where AI can do so much?

It means leaning harder into the things AI structurally cannot do. Not as a feel-good HR message. As a business strategy.

Here are five that actually matter in your customer experience:

1. Own the moments that carry real stakes. AI is excellent at low-stakes, high-volume interactions. It struggles when the interaction carries emotional weight. And your customers feel it. A billing dispute. A medical concern. A service failure that happened at the worst possible time. These are the moments that determine whether a customer stays or leaves. Put your best humans there.

2. Use AI to make your humans better, not fewer. The strongest operations I've seen don't use AI to replace agents. They use it to reduce the cognitive load on agents. Surfacing the right information faster. Suggesting next steps. Summarizing call history before a transfer. The agent still drives. The AI handles the prep work. The outcome is faster, more accurate, and more human. Not less.

3. Stop measuring containment. Start measuring resolution. Your AI vendor will show you how many calls didn't reach a human. That number tells you almost nothing about whether your customers got what they needed. Measure repeat contacts. Measure post-interaction NPS specifically on AI-handled calls. The gap between what your dashboard shows and what your customers actually experience is where brand damage quietly accumulates.

4. Make the path to a human visible and easy. A recent study found that more than half of American consumers actively try to circumvent chatbots. Some resort to yelling "human" at their phones. That's not a technology problem. That's a design problem. If your customers have to fight to reach a person, you're not saving money. You're just moving the frustration downstream. Always give them a door.

5. Your customers will tell you when you've gone too far. But by then the damage is done. Customer sentiment doesn't crash overnight. It erodes. CSAT drops a point. Repeat contacts tick up. Then a quarter later someone pulls the churn data and the numbers are ugly. By the time the signal is loud enough to act on, you've already lost customers you won't get back. Watch the leading indicators, not just the lagging ones.

The companies that will win in the next five years aren't the ones who automate the most.

They're the ones who figure out exactly where the human matters and protect that relentlessly, even when the spreadsheet says otherwise.

AI is not the threat to your customer experience. Forgetting why your customers chose you in the first place is.

That choice, in almost every case, had something to do with a person.

What’s your take? Tell me on LinkedIn. I read every comment.