Plain Answers·Published ·7 min read

AI chatbot for a small business: what it should do, and when to skip it

I built my first chatbot in 2018 for a sports bar in Burlington. The technology is far better now. The reasons chatbots fail haven't changed at all. Here's what to check before you put one on your website.

SD

Steffen deGraaf

Founder & Forward Deployed Engineer · Burlington, ON

Short answer: an AI chatbot is worth it for a small business when it has one clear job, a clean hand-off to a person, and someone who reads the conversations after launch. Without those three, it's a widget in the corner of your website that annoys people.

Let me ask you something.

When someone lands on your website at nine at night with a question, what happens? For most businesses the answer is nothing. They read a bit, they leave, and maybe they call tomorrow. Maybe they call your competitor.

That gap is what a chatbot is for. And it's a real gap. But I've been building these since 2018, and I've watched a lot of them get installed with real excitement and switched off three months later. So before you buy one, let's look at what actually makes the difference.

Quick Answer

Does a small business need an AI chatbot?

Only if the same questions keep arriving when nobody is there to answer them, and an answer or a booking is worth money to you. A good small-business chatbot does one job (answer the common questions, book the appointment, or capture the lead), hands off to a person with a summary when it's out of its depth, and gets tuned from real conversations in its first two weeks. If your questions are all different or all complicated, a phone number and a fast callback will serve you better.

  • →One job first: answer the common questions, book the appointment, or capture the lead
  • →Design the hand-off to a person before anything else
  • →Read the first two weeks of conversations and fix what they show you
  • →Say plainly that it's an AI assistant, and give it a consistent voice
  • →Skip it when every question is different or the stakes are high

What does an AI chatbot actually do for a small business?

Strip away the sales page and a chatbot does three useful things. Pick one to start.

  • Answers the questions you answer every day. Hours, service area, what's included, whether you take a certain kind of job. The stuff your front desk could recite in their sleep.
  • Books the appointment. If your calendar is connected, it can offer times and put the booking in without a phone call.
  • Captures the lead. Name, number, what they need, how urgent it is. Then it gets that to a person who follows up.

That third one is where the dollars are. A chatbot that takes a tidy message at nine at night and puts it in front of the right person at eight the next morning has done its job, even if it never answered a single question.

Here's what it shouldn't do. It shouldn't quote prices on work it can't see. It shouldn't make judgment calls. And it shouldn't pretend to be a person.

What makes a chatbot fail?

The same three things, every time. None of them is the technology.

1. It was asked to do too much

In 2019 I took on a chatbot for a dental clinic. The brief was simple: book appointments and answer common questions. By week three it was also supposed to handle insurance billing questions, refer to staff by name, and track loyalty points. By week six we had something too complicated to maintain and too brittle to trust.

I see the same pattern now with far better tools. Somebody says “let's add a support bot,” and two meetings later it's meant to process refunds and update the CRM as well. One job. Get that working with real customers. Then decide what's next from what they actually asked for.

2. Nobody designed the hand-off

The early bots I built either tried to handle everything and fell over on the odd cases, or offered to “connect you with a team member” the second things got interesting, which made the bot feel pointless.

The ones clients kept had a third option. The bot said, more or less: this one is outside what I can handle, so I'm passing it to the team with a summary so you don't have to repeat yourself, and here's when you'll hear back. Then a person actually got that summary and actually replied.

So before you choose a tool, answer four questions on paper. What triggers a hand-off? What information goes with it? Who receives it? How fast do they respond?

3. Nobody looked at it after launch

I've built systems that were perfect in testing and fell apart on day three. You can't predict what people will type. You can only watch.

For the first two weeks, read the conversations. How many did it resolve without a person? How many did it hand off? And read the transcripts of the ones that went badly, word for word. Those tell you exactly what to fix. Plan for that tuning time up front. It isn't polish. It's the part that makes the thing work.

“

The technical build is rarely the hard part. The hard part is building something that survives contact with real customers.

Should the chatbot say it's an AI?

Yes. Say what it is, plainly, and then make it good.

Back in 2018 we found that the moment people knew it was a bot, a lot of them assumed it was stupid and stopped typing. Bots that tried to pass as human were worse, because people figured it out and stopped trusting the business.

What worked was being honest about what it is and giving it a consistent voice. Warm, direct, a bit of personality, no fake name with a fake backstory. Today's AI models are very good at holding a voice. The work is in telling it clearly what tone to use and what to stay away from.

When should you skip the chatbot?

There's a really good reason to say no sometimes. Skip it when:

  • Your questions are all different. If no two enquiries look alike, there's nothing for the bot to learn.
  • The stakes are high on every conversation. Anything where a wrong answer costs a customer money or health belongs with a person.
  • Almost nobody visits the website. Fix that first.
  • Nobody will own it. A chatbot with no owner drifts out of date, and an out-of-date bot is worse than no bot.
  • Your customers call instead of typing. Then the missed call is the problem to fix, and an AI receptionist or a missed-call text-back is the better first step.

What should you check before you buy one?

  1. Write down the ten questions you get most. If you can't get to ten, you may not need a bot.
  2. Pick the one job: answer, book, or capture.
  3. Decide who gets the hand-off and how fast they reply.
  4. Ask the vendor to show you the conversation transcripts screen. If you can't read what customers typed, you can't improve it.
  5. Ask where the customer's messages are stored and who can see them, and read what to check about confidentiality before client details go through it.
  6. Put your own numbers through the ROI estimator. If the math doesn't work on paper it won't work on your website.

How does BotLogix build one?

The same way we do everything else. We start with one process. That means a free 30-minute AI Discovery Call to see whether a chatbot is even the right answer, then a $1,000 + HST on-site AI Strategy Day where we sit with the people who answer the questions today and map what the bot should handle and what it should pass along. The build is scoped after that, once we both know what it has to do.

We've been at this since that first Messenger bot in 2018. The full story, including the parts that went wrong, is in what eight years of building chatbots taught us.

Takeaway

A chatbot is not the strategy. Answering your customers faster is the strategy. If a bot does that for one clear job, hands off cleanly, and somebody reads the transcripts, it earns its place. If not, it's decoration.

If you'd like a second opinion on whether one fits your business, book the free 30-minute AI Discovery Call. Bring the ten questions you get most and we'll go from there.

Free 30-Minute AI Discovery Call

For real businesses with real problems.

Not for tire kickers. Not for tech tourists. If you have manual work, missed calls, or after-hours admin your team hates as much as you do — let's talk.

Book my Discovery Call
Steffen deGraaf

Written by

Steffen deGraaf

Founder & Forward Deployed Engineer · Building business systems since 2018

Eight years of shipping business systems. Based in Burlington, Ontario. Questions, pushback, or a war story to share?

Email me directly

Free 30-Minute AI Discovery Call

For real businesses with real problems.

Not for tire kickers. Not for tech tourists looking for a demo. For business owners who already know something in their operation is broken — and want to know if AI can fix it.

This isn't about replacing people. It's about killing the bad business processes your team hates as much as you do — the manual work, the missed calls, the after-hours admin.

Start with a free 30-minute AI Discovery Call. The AI Strategy Day is $1,000 + HST, capped at 4 people from your team. Larger teams or multi-location workshops by special arrangement — ask when you book.

Steffen deGraaf, founder of BotLogix

Your session is with

Steffen deGraaf

Founder, BotLogix · Burlington, ON