A Chatbot Is a Front Door, Not the Whole House

There's a version of customer support that's becoming popular and quietly making people miserable: route everything through the AI, and hope nobody notices there's no human behind it.

You've probably hit it. Some companies have gone all-in — nearly every support interaction runs through their AI assistant, and there's no obvious way around it. When the bot can handle your problem, it's fast and it's fine. But the moment your situation doesn't fit the script, you're stuck. You can't reason with it. You can't explain the nuance. You just keep rephrasing the same thing into a wall that keeps giving you the same answer.

I'm not here to bash any specific company for doing this. The tools are genuinely useful and the instinct to use them makes sense. My point is narrower: going all the way from "humans answer everything" to "the AI answers everything" is too drastic a jump. The right answer is in the middle.

What chatbots are actually good at

Let's be fair to the technology, because it earns its place. A chatbot is excellent at the high-volume, low-complexity end of support. General questions. "Where's my order?" "How do I reset my password?" "What are your hours?" "Can I change my plan?" These are the questions that make up the bulk of any support queue, and a bot answers them instantly, at 3 a.m., without a ticket, without a wait. That's a real win — for the customer and for the business.

For that layer, a chatbot isn't a downgrade. It's an upgrade. Nobody wants to wait on hold to ask a question the bot could have answered in two seconds.

Where it falls apart

The problem starts the moment a question gets specific.

Ask a chatbot something genuinely particular to your situation — an edge case, a billing dispute with context, a problem that needs someone to actually understand rather than match — and it struggles. And this is worse than the demos suggest, because the model running most production chatbots isn't the frontier, top-of-the-line one. It's a cheaper, lower-reasoning model, chosen precisely because it has to run at scale on every conversation. Those models are fine at retrieving an answer that already exists. They're weak at reasoning through a situation that doesn't.

So the customer with the non-standard problem — the one who most needs help — gets the least capable helper. They explain, they clarify, they repeat, and the bot keeps missing the point. That's where the human essence gets lost. Not in the easy questions, but in the hard ones, where being understood is the entire point of asking.

The answer is a mix, not a replacement

So here's what I actually believe works: a chatbot for general questions and quick resolutions, backed by a real human line for everything else.

Let the bot take the first pass. It resolves the common stuff instantly, which is most of the volume, and everyone's happy. But when it hits something it can't resolve — or something that clearly needs a human touch — it hands off cleanly to a person. That handoff can be a live chat, or a follow-up by email afterward. The channel matters less than the principle: there is always a way to reach a human, and the bot knows when to step aside.

This isn't a compromise between two worlds. It's the best of both. The customer gets speed on the easy things and understanding on the hard things. The business handles more volume without drowning its support team in password resets, and keeps its humans focused on the conversations that actually need a human. Nobody gets trapped talking to a wall.

Don't automate away the reason people trust you

Support isn't only about resolving tickets. It's one of the few moments a customer actually talks to your company, and how that feels is part of what they're buying. Automate the whole thing and you save money on paper while quietly spending down something harder to measure — the sense that there's someone there who cares when things go wrong.

A chatbot is the front door. It should be fast, welcoming, and able to handle most visitors on its own. But a front door isn't the whole house. Behind it, there still has to be someone home.