This is an opinion piece. The views are the author’s own.
I have a new hobby. It’s called typing “agent” into a chat window over and over until something changes. Sometimes it works. More often, a cheerful bot tells me it understands my frustration, offers me three help articles I have already read and asks whether I would like to “start over.”
I would not. I would like to talk to a person.
Let me be clear about what I’m not saying. I’m not against customer service chatbots. When I want to know a store’s hours, track a package or reset a password, a bot that answers in five seconds beats a 20-minute hold. Fine. Automate that. My argument is narrower and, I think, hard to disagree with: every chatbot that stands between a customer and a company should have a clearly labeled, always-visible way to reach a human. Not hidden behind a magic phrase. Not available only after you fail three times. A button.
Customers Have Already Said What They Want
This isn’t just me being grumpy, although I am. When Gartner surveyed 5,728 customers in December 2023, 64% said they would prefer companies not use AI for customer service at all, and 53% said they would consider switching to a competitor if they found out a company was doing so. Their top worry wasn’t robots taking jobs or bots making things up, though both made the list. It was that AI would make it harder to reach a human agent.
That’s a remarkably specific complaint. People aren’t asking companies to tear the bots out. They’re asking not to be trapped by them.
Federal regulators noticed the same thing early. In a 2023 report on chatbots in banking, the Consumer Financial Protection Bureau estimated that 37% of the U.S. population had interacted with a bank’s chatbot in 2022, and that all of the top ten commercial banks used them. The bureau warned about customers getting stuck in “doom loops” of repetitive, unhelpful replies with no way out to a person. It also pointed out who gets hurt most when human help disappears: people with limited technical skills or limited English.
Those are exactly the customers least likely to know that typing “representative” six times is the secret handshake.
Bots Are Bad at the Problems That Matter Most
Here is the part companies don’t like to admit. The questions a bot handles well are the ones you could have answered with a decent FAQ page. The reason people contact customer service is usually that something has gone wrong in a way nobody planned for: a double charge, a canceled flight with a connecting reservation, a delivery marked “received” that never arrived. These are messy, one-off problems, and they are precisely where a scripted system runs out of script.
The CFPB made a related point that sticks with me: automated systems may be less willing than people to do the things that actually fix a problem, like waiving a fee or negotiating. A human agent can look at a situation and decide that the rule doesn’t fit. A bot can only apologize in a loop.
Even the Biggest AI Booster Blinked
If you want a case study, look at Klarna, the buy-now-pay-later company. In early 2024 it said its AI assistant was doing the work of 700 human agents. It was the poster child for replacing people with bots. Then, in May 2025, CEO Sebastian Siemiatkowski told Bloomberg that cost had become too dominant a factor and the result was lower-quality support. Klarna started hiring human agents again so customers would always have the option of reaching a person, while saying it remains an AI-first company.
Good for them for saying it out loud. But most companies won’t have that moment of public reflection. They’ll quietly shrink the phone team, bury the number and count fewer “contacts” as a win. Fewer contacts is not the same as fewer problems. Sometimes it just means more people gave up.
What a Fair Chatbot Looks Like
None of this requires banning anything. It requires a few basic manners:
- A visible human option from the first screen. A button or menu item, not a password you have to guess.
- An honest wait time. If a person won’t be available for 40 minutes, say so, and offer a callback.
- A handoff that keeps the context. If I’ve already explained my problem to the bot, the human should see that transcript. Making me start from zero is a second penalty.
- A clear label. Tell me when I’m talking to software. I’m an adult; I can handle it.
Companies that do this will find that most people still use the bot for the easy stuff, because the easy stuff is genuinely faster that way. They’ll just stop feeling like hostages when it isn’t.
Automation should be a shortcut customers choose, not a wall they have to climb. Put the button in. It costs less than losing the customer, and according to Gartner’s numbers, plenty of them are ready to walk.
More on Contoh
- No, I Will Not Rate My Experience
- Stop Making Me Download an App to Park
- Your Smart TV Should Ask Before It Watches What You Watch
Sources
- Gartner: Survey Finds 64% of Customers Would Prefer That Companies Didn’t Use AI for Customer Service
- CFPB: Issue Spotlight Analyzes “Artificial Intelligence” Chatbots in Banking
- Consumer Finance Monitor: New CFPB report highlights concerns with growing use of chatbots by financial institutions
- Fortune: Klarna’s CEO on rehiring humans for customer service