Every company wants to tell you about their amazing new AI customer service assistant. It’s available 24/7. It can handle thousands of queries simultaneously. It never gets tired or cranky.

What they don’t mention is that most customers would rather chew glass than interact with it.

I’m not exaggerating. Recent surveys show that over 60% of customers prefer talking to a human, even if it means waiting longer. The enthusiasm gap between companies deploying AI customer service and customers actually using it is massive.

What’s Going Wrong

The problem isn’t AI itself. It’s how it’s being implemented. Most companies are using AI as a cost-cutting measure disguised as an innovation. They’re replacing humans with bots, not enhancing the customer experience.

You’ve experienced this. You have a specific question about your account. The chatbot gives you three generic options that don’t match your issue. You try to explain. It doesn’t understand. You ask for a human. It insists it can help. You get frustrated. Eventually you find a way to escalate, and by the time you reach an actual person, you’re already annoyed.

That’s not the AI’s fault, exactly. It’s doing what it was designed to do: deflect as many queries as possible away from expensive human agents. The company saves money. You waste time. Everyone loses except the spreadsheet.

When AI Customer Service Actually Works

There are exceptions. Some companies are getting this right. The difference? They’re using AI to augment human service, not replace it.

Good AI customer service handles the genuinely simple stuff. Password resets. Order tracking. Basic FAQs. Things with clear inputs and clear outputs. It does these instantly, which is great for customers who just want a quick answer.

But it also knows when to get out of the way. If the query is complex or emotional or unusual, it hands off to a human immediately. No runaround. No “let me try to help you with that” when it clearly can’t.

I’ve seen a firm that does this well implement systems where AI handles tier-one queries and enriches the handoff with context. When a human agent picks up, they already have all the relevant information. The customer doesn’t have to repeat themselves. That’s using technology to make the human interaction better, not to avoid it.

The Transparency Problem

Here’s something customers hate: not knowing if they’re talking to AI or a human. Companies often make their bots sound human-like, with names and personalities. They think this makes the experience better. It doesn’t.

People can tell. The uncanny valley effect is real. When you think you’re talking to a person and then realise it’s a bot, it feels deceptive. Trust takes a hit.

Better approach? Be upfront. “Hi, I’m an AI assistant. I can help with X, Y, and Z. For anything else, I’ll connect you with a team member.” Clear. Honest. Respectful.

The Pattern Matching Trap

AI is brilliant at pattern matching. It’s terrible at understanding context. This creates problems.

A customer writes: “My package was supposed to arrive yesterday for my daughter’s birthday. Now she’s upset and I need to know where it is.”

A bad AI system sees this as a tracking query and spits out a tracking number and estimated delivery date.

A human sees that this is about a disappointed kid and a stressed parent. The response needs empathy, not just data. Maybe expedited shipping. Maybe a small discount. Maybe just an acknowledgment that this situation sucks.

This is why customer service will always need humans. Not for every query, but for the ones that matter. The ones where emotion and context are as important as information.

What Customers Want

The data is pretty clear on this. Customers don’t hate AI. They hate bad experiences. If AI gives them what they need quickly, they’re fine with it. If it wastes their time and makes them jump through hoops, they’re not.

What they really want is options. Sometimes you want to self-serve. You’ve got a simple question, it’s 11pm, you just want an answer. An AI that can genuinely help is great for that.

Other times you want a human. Your issue is complicated or you’re frustrated or you just prefer talking to a person. That should be easy to access, not hidden behind layers of bot deflection.

The Future Isn’t Botless or Humanless

We’re going to see more AI in customer service, not less. That’s inevitable. But the companies that get it right will be the ones that use AI thoughtfully.

That means AI for speed and scale on simple queries. Humans for complexity and empathy. Clear handoffs between the two. Transparency about what you’re interacting with.

It also means measuring the right things. Not just “queries deflected” or “cost per interaction.” Those metrics optimise for the company, not the customer. Better metrics: resolution time, customer satisfaction, first-contact resolution.

AI customer service can be good. It just rarely is, because companies are optimising for the wrong outcomes. The ones that figure this out will have a real competitive advantage.

The rest will keep annoying their customers while wondering why satisfaction scores keep dropping.