What a website chatbot actually does for a one-truck shop
A chatbot is a small chat window in the corner of your site that answers questions the second someone lands on the page, at 11pm on a Tuesday or during a Sunday storm when your phone's off. It's not a phone-answering AI (see the guide on ai-receptionist-basics at /guides/ai-receptionist-basics), it's for the visitor who'd rather not call yet. A chatbot catches that person before they close the tab and forget you exist.
Done right, it answers the questions you get asked every week (do you serve my town, what's a service call cost, can you come this week), then grabs a name and number so you can follow up. Done badly, it's generic text that frustrates people into leaving. The gap comes down to how much real information you put into it, which is most of this guide.
Rank the options: free, cheap, and paid
Free tier: Tidio and Tawk.to both have real free plans, not trials. Tawk.to is free for a live-chat widget with canned responses and unlimited chats, no card required. Tidio's free tier caps out at a small number of conversations a month (check the current limit at signup) with simple rule-based replies: type 'pricing,' get a canned answer.
Cheap AI tier, roughly $20 to $50 a month as of mid-2026: Tidio's paid plans add an AI assistant (branded Lyro) that reads pages you feed it and answers in real sentences instead of buttons. Intercom and Crisp sit in a similar range. This tier is the sweet spot for most service businesses.
Full AI tier, roughly $50 to $150+ a month: tools built for lead capture, like Podium's AI website chat or ServiceTitan's chat add-on if you already run their field service software, connect the bot straight into your job scheduling and text follow-up.
Done-for-you: paying someone to build, train, and maintain the whole thing, wired into your CRM from day one. That's the paid path, worth naming honestly since it's real money for real time saved.
Pick one and set it up (30 to 45 minutes)
Start with Tawk.to (free) or Tidio's free tier. Both work the same way: create an account, verify your email, land on a setup wizard.
Tawk.to generates a small snippet of code, one script tag. If your site runs on Squarespace, Wix, or a similar builder, look for "code injection" or "custom code" in site settings and paste it there. If someone else built your site, send them the snippet and ask them to drop it before the closing </body> tag. Both tools have one-click plugins for WordPress and Shopify too.
Set your business hours in the widget settings so it honestly shows "we're open" or "we're closed, leave a message." Nothing kills trust like a chat bubble promising a live person at 2am.
Train it on your own FAQ, not a generic script
This is the step people skip and it's the whole game. Write down the 8 to 12 questions you actually get asked, word for word if you can. Not "tell me about your services," the real phrasing: "do you come out to Hartland," "how much for a service call."
Write a short, honest answer for each, 1 to 3 sentences, the way you'd say it to a neighbor. Then load those pairs into the tool. Tawk.to calls this Shortcuts, Tidio's paid tier calls it training the assistant on your content.
A few that belong on almost every contractor's list:
- "What areas do you serve?" (name the actual towns, not "the greater area")
- "What does a service call cost?" (a real number or range, even if it's "$89 plus parts")
- "Are you licensed and insured?" (yes, and say it plainly)
- "How fast can you come out?" (be honest about your actual lead time)
- "Do you offer financing?" (yes or no, don't dodge it)
- "Do you work weekends?"
If your site already has good FAQ content, the chatbot's job gets easier. Pages written to answer real questions (see the guide on content-that-ranks-locally at /guides/content-that-ranks-locally) give an AI-powered bot real material to pull from.
Build the handoff to a human before you launch
Every one of these tools has a mobile app or notification that pings you the moment a real visitor starts chatting, and lets you jump in live. Turn notifications on before you turn the widget on. A live quote or an upset customer needs you, not a script.
Set a trigger for when the bot should stop guessing and hand off. If the visitor asks about a specific problem at their house (a leak, a smell, a noise, an error code), the bot's reply should be some version of "that's worth a real conversation, what's the best number to reach you." Most platforms let you set this as a fallback for anything outside your trained Q&A.
Whatever the bot collects, make sure it lands somewhere you actually check daily. A lead sitting unread in a dashboard for four days did nothing you couldn't have done with no chatbot at all.
What a bad bot costs you, compared to no bot
No chatbot costs you the after-hours visitor who leaves with no way to reach you and never comes back, but that's a cost you can't point to. A bad chatbot is a different cost: a visitor asks a real question, gets a wrong price, and now they've had a worse experience than if the bubble hadn't been there. That's what happens when you turn one on and never train it.
The fix: revisit the Q&A pairs every few months, especially after a price change or a new service. Read a week of chat transcripts once a month, most tools save them, and you'll spot what your bot fumbled.