There was a time when calling customer support meant talking to a machine.
"Press 1 for billing. Press 2 for technical support. Press 3 for something that sounds vaguely related to your problem."
Somewhere, hidden deep inside this telephone labyrinth, there was a human being. Experienced customers knew the trick: say "customer representative," press zero repeatedly, or shout "OPERATOR!" until the system finally surrendered.
It wasn't a great system. But at least there was an exit.
Then came AI. And somehow, we managed to make it worse.
Meet the New Wall
Today, more and more companies are putting AI assistants in front of their customer support. The promise sounds wonderful: available 24/7, instant responses, intelligent conversations, personalized assistance.
And of course, significantly lower operating costs. Especially that last part.
Instead of having people handle customer requests, companies can deploy an AI assistant trained on their documentation and policies. It answers politely, never gets tired, never asks for a salary increase, and can apologize in seventeen languages.
And when it cannot solve your problem, it can apologize again. Beautifully.
The old telephone system had a limited number of ways to tell you it couldn't help. AI has an almost unlimited vocabulary for doing exactly the same thing.
The Most Polite Way to Say No
We recently experienced this ourselves. A virtual server had been booked by mistake, and we wanted to cancel it. Nothing particularly complicated. Just cancel a server we didn't need.
So we contacted support. Or rather, the AI that stood between us and support.
We explained the situation, and the AI explained the policy. We clarified the mistake, and it explained the policy differently. We asked for an exception, and it expressed understanding before explaining the policy again.
When we explicitly asked to speak to a human, the AI wouldn't let us through. And we couldn't find another way to contact anyone.
The assistant was extremely well grounded. It knew exactly what it was allowed to say and what it wasn't allowed to do. Unfortunately, solving our problem wasn't part of its mission.
We were negotiating with a very polite wall.
Eventually, we disputed the credit card payment. And suddenly, we got a reaction. Apparently, the payment system had a better escalation procedure than customer support.
When Cost Reduction Becomes the Product
Let's be clear: there is nothing wrong with using AI in customer support. Quite the opposite. AI can find answers, analyze technical logs, identify known problems, suggest solutions, and help support engineers respond faster. It can eliminate repetitive work and free people to handle difficult cases.
That's real progress.
But there is a fundamental difference between using AI to improve customer support and using AI to avoid providing customer support.
One starts with the customer. The other starts with a spreadsheet.
If the primary objective is reducing costs, the easiest solution is obvious: make it harder for customers to reach expensive human employees. Put an AI assistant in front of them, let it handle simple questions, and repeat company policies for everything else.
Then measure how many conversations never reach a human and call that efficiency. The customer who gives up after twenty minutes? Another successfully deflected support ticket.
Congratulations. The dashboard looks fantastic.
The problem isn't the AI itself. It's the mission it has been given. An assistant should help people get things done, not become a gatekeeper preventing them from reaching someone who can.
Even a perfectly trained AI can deliver terrible customer service if it has been designed around the wrong objective. In fact, better AI may make things worse: the more convincing it becomes, the longer it can keep customers trapped in conversations that lead nowhere.
The old telephone menu couldn't pretend to understand your frustration. The new AI can. That doesn't mean it can do anything about it.
At Ice Tea Group, We Use AI. A Lot.
We are certainly not against AI. At Ice Tea Group, it has become an important part of how we work. We use it heavily in software development and across our daily processes. It helps us move faster, investigate problems, improve quality, and reduce repetitive work.
And we're only getting started.
But there is one thing we don't want to forget: we work for our customers, not the other way around.
Our customers don't contact us because they enjoy opening support tickets. They contact us because they need something. Sometimes it's a simple question, sometimes a complex technical problem, and sometimes something went wrong or someone made a mistake.
And sometimes the answer simply isn't in the documentation. That's precisely when support matters most.
AI should help us understand problems faster, suggest solutions, and make our people more effective. It shouldn't become an excuse to stop listening. We don't want AI to stand between us and our customers. We want AI to help us serve them better.
Don't Build a Better Wall
We have AI systems capable of understanding complex requests, reasoning through technical problems, and accomplishing tasks that were difficult to automate just a few years ago.
And what are some companies doing with this extraordinary technology? Building a more sophisticated version of "Press 1 for billing." Except now you can't press zero to escape.
There is a better approach: give AI the knowledge and tools to solve what it can, and when a problem requires human judgment, flexibility, or authority, make the transition effortless.
Because the objective of customer support is to support the customer, not to minimize the number of people who manage to reach it.
If AI in customer service is primarily a cost-cutting exercise, designed to replace people without improving the service, it is doomed to disappoint. It may reduce expenses and improve internal metrics, but customers will remember something else.
They asked for help. And the company built a wall.
The future of customer support shouldn't be AI instead of people. It should be people doing a better job because of AI.