Most Chatbots Annoy People. Here Is What Makes a Good One

We have all rage-clicked through a chatbot that just would not understand a simple question. So when a client says "we want a chatbot," our first reply is usually: okay, but let us make it one people actually like.

The bad ones are not bad because the technology is weak. Language models today can hold a perfectly reasonable conversation. They are bad because of decisions made during planning, or skipped entirely. Here are the mistakes we see most often, and what a good chatbot does instead.

Mistake 1: Starting with the tool instead of the questions

Many projects begin with "which platform should we use?" before anyone has looked at what customers actually ask. The result is a bot that is technically impressive and practically useless.

Do this instead: pull a month or two of real conversations from your inbox, live chat, WhatsApp and support tickets. Group them by topic. You will usually find that a small set of questions makes up most of the volume: order status, pricing, opening hours, how to reset a password, refund policy. Those are your first version.

Mistake 2: Forcing people through rigid menus

"Press 1 for billing, 2 for technical support" belongs on phone lines from twenty years ago. When a customer types "my package hasn't arrived and it's been a week," they should not be pushed back to a menu of six buttons that do not match their problem.

Do this instead: let people type in plain language, including typos, slang and mixed languages, and use buttons only as shortcuts for common next steps.

Mistake 3: Letting the bot make things up

A general-purpose AI model will confidently answer questions about your business using guesses. That is how a bot ends up promising a discount that does not exist or quoting a policy you dropped last year.

Do this instead: ground the bot in your own material: your real FAQs, product catalog, pricing, policies and terms. Instruct it to answer only from that material and to say "I'm not sure, let me get someone" when the answer is not there. This is the single biggest difference between a bot people trust and one they screenshot and mock on social media.

Mistake 4: Trapping people in a loop

Nothing is more infuriating than a bot that responds to every frustrated message with "Sorry, I didn't understand. Here are some topics I can help with."

Do this instead: set clear triggers for handing over to a human. Two failed attempts, an explicit request for a person, an angry tone, or a sensitive topic such as a complaint or a payment problem should all route to your team. And if nobody is available, say so honestly and tell the customer when they will hear back.

Mistake 5: Making customers repeat themselves

A handover that dumps the customer into a new queue where an agent asks "How can I help you today?" undoes all the goodwill the bot earned.

Do this instead: pass the full conversation and the details already collected (name, order number, issue) to the human agent, so they can pick up mid-sentence.

Mistake 6: A bot that can only talk

If the bot can explain your refund policy but cannot tell a customer where their order is, it is a slightly friendlier FAQ page.

Do this instead: connect it to the systems that hold the answers: order management, booking calendar, CRM, inventory. This is usually where most of the real work in a chatbot project sits, and it is what makes the bot genuinely useful. Plan for it from the start with proper API integration, not as a phase two that never happens.

Mistake 7: Putting it in the wrong place

A beautiful chatbot on your website does not help much if your customers mostly message you somewhere else. In Indonesia, that somewhere else is very often WhatsApp, where a WhatsApp AI chatbot can meet people in the chat they already use every day.

Do this instead: launch on the channel with the most volume first, then expand.

Mistake 8: Launch and forget

Your products, prices and policies change. If the bot's knowledge does not, it starts giving wrong answers within weeks.

Do this instead: assign an owner. For the first month, review conversation logs every week: where did the bot fail, what did people ask that it could not answer, where did handovers happen? Update the knowledge base and adjust. After that, a monthly review is usually enough.

Before you launch, run this quick test

• Ask your top ten questions in three different ways each, including one with typos.
• Ask something it should not know and confirm it admits that.
• Ask to speak to a human and time how long the handover takes.
• Try to trick it into offering a discount or promising something off-policy.
• Have someone outside the project use it cold and watch where they get stuck.

When you probably do not need a chatbot

If you get a handful of messages a day, a well-organized inbox and quick replies from your team will serve customers better. The same goes for sales that depend on long, consultative conversations where every case is different. A chatbot pays off when the same questions arrive over and over, at all hours, faster than your team can comfortably answer them.

Done right, a chatbot quietly handles the repetitive questions so your team can focus on the conversations that need a human. If you want one built around your business instead of a generic template, this is our AI chatbot development service. Start with your top ten customer questions and grow from there.