A few years ago, every client wanted the same thing: a homepage carousel. That rotating strip of images that moves on its own and nobody looks past the first slide. Then it was cookie banners on websites with no cookies. Then Google Tag Manager in businesses that never opened a single analytics report.
Now it’s the AI chatbot.
A developer talked about this in a post that went viral on Hacker News this week, and it’s worth pausing to think about what he described. When clients ask him for a chatbot, he asks them a simple question: "Do you use chatbots when you visit other websites?" There’s usually a pause. Then a laugh. No, honestly, not really. They close them immediately. They find them annoying.
And yet, everyone wants one on their own site.
This isn’t a criticism of AI or chatbots. It’s a criticism of automating because it’s trendy instead of because it’s necessary. And it’s exactly the conversation we have every week with clients who arrive saying "I want a bot" without having thought about what problem they’re actually solving.
The pattern repeats every 3-5 years
Look at the recent history of web design and you’ll see the same cycle again and again:
| Year | The thing everyone "needed" | Real result |
|---|---|---|
| ~2012 | Image carousels | Nobody gets past the first slide |
| ~2015 | Cookie banners | Everyone clicks "accept" without reading |
| ~2018 | Google Tag Manager | Installed and never used |
| ~2020 | Newsletter pop-ups | Closed before being read |
| ~2025 | AI chatbots | Closed or ignored |
The mechanism is always the same: a big player does it, everyone else copies it, and suddenly your business seems to "need" that thing to be taken seriously. The carousel didn’t die because someone decided it was bad. Something newer simply came along to copy.
What’s worrying about chatbots is that when they’re badly built, they’re not just decorative: they’re actively harmful. A client told this developer that a competitor’s chatbot had spent months confidently giving out wrong opening hours. That’s not harmless decoration. That’s lost customers.
Why so many chatbots are useless (and how to spot it)
A poorly thought-out chatbot has three clear symptoms:
1. It knows nothing about your specific business. It’s a ChatGPT wrapper with a logo slapped on top. If you ask about opening hours, prices, or parking, it hallucinates or makes things up.
2. It’s in the wrong place. On an informational five-page website, a chatbot is noise. The user wants to read, not chat with a robot that will go in circles to answer something already in the menu.
3. Nobody measures anything. Just like the client who forgot their Google Analytics login, there are businesses with chatbots that have been live for months and they have no idea how many conversations have happened, how many turned into sales, or what users are actually asking.
If your "AI automation" has these three symptoms, it’s not automation. It’s expensive decoration.
When a bot does make sense — and when it doesn’t
The right question isn’t "Do I want a chatbot?" It’s "What repetitive task is costing me time or money today?"
Examples where a well-built bot delivers real, measurable value:
- You receive 30+ calls a week outside business hours and lose leads because nobody answers. In that case, a voice bot that takes messages and books meetings pays for itself in weeks.
- Your team spends 2-3 hours a day sorting and answering repetitive emails. An automated inbox gives you that time back in full.
- Sales reps call unqualified leads and waste time on people who were never going to buy. An automatic qualification system enriches each lead before it reaches the team.
- You have an FAQ that people can’t find and customer support answers the same thing 50 times a day. Here, yes, a chat bot trained on the business’s real documentation reduces the load.
Examples where it does NOT make sense:
- Your site is a five-page brochure and gets 200 visits a month.
- You run a local business with a phone you already answer just fine.
- You don’t know what problem you want to solve, but "everyone has one."
The simple rule: start with the problem, not the tool
The line we repeat to new clients is this: describe the problem, not the solution. If you come in saying "I want a chatbot," the first thing we do is turn the conversation around. What specific task is eating up your time? How many hours a week? What happens if it doesn’t get done? How much does it cost not to do it?
Sometimes the answer is a bot. Sometimes it’s a quiet workflow that connects your CRM to your email and nobody ever sees — but it saves you four hours a day. Sometimes, honestly, it’s doing nothing and putting that budget somewhere else.
This goes against the instinct to "look modern." But there’s an uncomfortable observation in the original post that’s worth quoting: when a client says "simple," they often don’t mean easy to use. They mean not impressive. They mean what will other people think. A fast, clean website doesn’t look like it cost anything. It doesn’t show effort. It doesn’t say: we take this seriously.
The chatbot is often exactly that: a "look, we’re modern" badge stuck in the corner of the homepage. It doesn’t solve anything. It just signals.
What actually works: invisible automation
The best automations we build are the ones the end customer never sees. An email that gets answered automatically in under a minute when it used to take two hours. A lead that reaches sales already enriched with LinkedIn data. A sales proposal generated in 10 minutes instead of two days. A workflow that doesn’t break on weekends or when an API changes.
None of this has a shiny widget in the corner of the website. And precisely because of that, it works. Building something genuinely simple, that loads instantly and says exactly what it needs to say and nothing more, is often harder than just dropping in a chatbot.
The same applies to automation. A bot blinking in the corner is the easy option. A system that truly saves your team 4 hours a day, integrated with your real tools, without breaking — that takes thought.
How to decide, in five questions
Before spending a euro on AI automation, ask yourself this:
- What specific task do I want to eliminate? If you can’t describe it in one sentence, you’re not ready yet.
- How much time does that task cost me today? In hours per week. If it’s under 5 hours, it probably isn’t worth it.
- What happens when that task doesn’t get done? You lose sales, lose customers, pay overtime, miss deadlines. Quantify it.
- Who will maintain this if it breaks? An automation without maintenance becomes a problem in 6 months.
- What success signal will I measure after 30 days? If you don’t have one, you’ll end up like the person with Google Analytics who can’t remember the login.
If you have clear answers to those five questions, automation makes sense. If you don’t, what you need isn’t a bot. It’s an honest conversation about how your business actually operates.
The conclusion
AI is not the carousel. AI is genuinely changing how businesses work, and those who don’t use it well will fall behind. But "using it well" doesn’t mean slapping a trendy widget on your website. It means identifying specific repetitive work and removing it at the root.
The difference between useful automation and decorative junk isn’t the technology. It’s whether someone bothered to ask, "what problem are we solving here?" before building it.
That question is free. And it should be asked first, not last.