AI chatbots for small local businesses: 24/7 lead capture

A physiotherapy clinic in Lyon was getting 40-50 site visits a day but only 3-4 contact-form submissions a week. Most visitors bounced on the pricing page — they wanted a quick answer about insurance coverage and were not going to wait for a callback to get it. We replaced the contact form with an AI chatbot trained on their FAQ, services, and insurance policies. Two months in, the clinic was capturing 28-32 qualified leads a week. Same traffic, same website, same clinic. The difference was a chatbot that answered at 10pm on a Sunday when humans could not.

This is the one piece of AI that pays for itself fastest. €19/month of tooling replaces what used to require a receptionist staffed to a web chat window. The trick is not installing it — anyone can drop a widget on a site in 20 minutes. The trick is setting it up so prospects get useful answers, high-intent leads get pushed into your CRM, and edge cases escalate to you instead of dying in a bot loop.

What a chatbot should actually do for a local business

Most chatbot advice is written by SaaS marketers. Their bots do "conversational commerce", "lead scoring", and "omnichannel engagement". For a local business, ignore all of that. Your chatbot has three jobs, and if it does them well you have the best ROI tool in your stack.

Job one: answer the common questions you get every day. Prices, hours, parking, insurance, "do you do X?", "how long does Y take?". Every one of those that lands in a human's inbox is wasted time. Job two: qualify leads and book them into your calendar. Not just collect names — actually push qualified prospects into your Cal.com or booking page, and dump the unqualified ones into a CRM with a tag so you can ignore them without losing them. Job three: escalate to a human when needed. Angry customer, urgent medical question, service you do not offer — all need to exit the bot cleanly.

Installing a chatbot that only does job one is a basic FAQ page with extra steps. Installing all three is what makes it a lead-capture system. This sits inside the broader AI lead generation playbook — the chatbot is the capture layer.

System 1 — Picking the right platform

We have tested Chatbase, Tidio, Intercom, Drift, ManyChat, and a stack of smaller players across client deployments. Our default for local SMBs is Chatbase — €19/month, trained on documents you upload (your site, service PDFs, FAQ), embeds in one script tag, and its API handles the booking webhook cleanly. It does not try to sell you features you do not need.

What to do: create a Chatbase account. Upload your website as a crawl source, plus 2-3 PDFs (services, pricing, FAQ). Let it index for 5 minutes. Test with 10 questions a real customer might ask. If any answer is wrong, edit the source document rather than the bot — cleaner long-term. Embed the script on every page of your site.

Tidio at €29/month is the alternative if you want human handoff built in (the owner can jump into a live chat from an app). Skip Intercom (€74/month minimum, designed for enterprise) and ManyChat (Facebook-first, awkward for service sites). Avoid free chatbots without document training — they hallucinate on your own services.

System 2 — Writing the prompt that stops it sounding robotic

Default prompts are where most chatbots fail. "I am a helpful assistant for [Business Name]" produces the generic corporate-speak that customers recognise instantly. Replace it with a 300-word custom prompt that covers four things.

What to do: write the prompt as if you were briefing a new junior receptionist on their first day. Include: your business's tone (formal? friendly? casual?), your core value proposition in one sentence, 3-4 things the bot should never say (avoid medical diagnosis, avoid quoting exact prices without context, do not promise same-day availability), and the escalation trigger (when to say "I will have someone call you back" instead of answering).

Measured outcome: a beauty salon in Nice rewrote their chatbot prompt from the Chatbase default to a 280-word brief that included "speak casually like a friend", "never push upsells on the first message", and "mention our studio's plants if asked what the space is like". Conversation completion rate went from 42% to 71% over four weeks. Booked appointments from chat went from 3/week to 9/week.

Want us to write the prompt, wire the booking, and deploy a production-ready chatbot for your business? Book a free 30-minute audit — we install it and hand over the account to you.

System 3 — Routing leads to your CRM and calendar

A chatbot that only answers questions is an FAQ. A chatbot that pushes qualified leads into your CRM and books them into your calendar is a growth engine. The wiring takes an hour and pays for itself in the first week.

What to do: decide what makes a lead "qualified" in your business. For a dentist: they want a specific treatment and can come in within two weeks. For a plumber: they have a real address in your service area and an actual issue. In Chatbase, create a "Lead form" that triggers after the bot detects those signals. The form collects name, phone, and intent in 3 fields.

On submission, Chatbase fires a webhook. Wire it through n8n (or Make if you prefer): branch on qualification, push qualified leads to HubSpot Free with a tag, and send them a Cal.com link via SMS through Twilio. Dump unqualified leads to a separate HubSpot list for periodic review. Send yourself a Slack DM for high-intent ones. This takes 45 minutes to build once. It runs forever. For the handoff from chatbot into your CRM sequences, see the admin-automation playbook.

System 4 — Reviewing conversations weekly to improve

Chatbots are not set-and-forget. The businesses that get 4× the conversions from the same bot are the ones reviewing logs every week. Chatbase and Tidio both show a conversation-log view. The review takes 30 minutes per week and is the highest-leverage time you will spend.

What to do: open last week's conversations. Skim for two patterns. First pattern: a question the bot failed to answer or answered wrong. Fix by updating the source document (never by editing the bot's mid-prompt — that gets messy). Second pattern: a conversation where the bot should have escalated but did not. Fix by tightening the escalation trigger in the prompt. Over 6-8 weeks of this, the bot gets dramatically better.

A dental clinic we worked with in Bordeaux caught 11 "failed" conversations in month one — all edge cases about dental implants, a service they offered but had not documented. One afternoon adding an implants FAQ PDF to Chatbase removed the issue. Booked consultations from chat for implants jumped from 0 to 4 per month.

Frequently asked questions

What is the best AI chatbot for a small local business in 2026?

For most service SMBs: Chatbase at €19/month. Trained on your own documents, embeds in one script tag, handles booking handoff via webhook. Tidio is the runner-up if you also need human handoff. Skip Intercom and Drift — overbuilt and overpriced. Skip free options without document training.

Will an AI chatbot sound robotic to my customers?

Only if you accept the default prompt. A 300-word custom prompt that specifies your voice, what to avoid saying, and what to escalate makes a chatbot indistinguishable from a junior employee for the first three exchanges. Past that, handoff to a human is the right move.

How long does it take to set up an AI chatbot?

From zero to live: about four hours for a solo owner comfortable with web tools. Sign-up, upload documents, write the prompt, embed the script, connect the booking webhook, and test. Weekly maintenance: 30 minutes reviewing conversations.