Should Your Small Business Have an AI Chatbot? An Honest Answer
I get asked about AI chatbots constantly now. Every vendor demo ends with the little chat bubble. The honest answer: for most small businesses, the chatbot is the last thing to add, not the first.
Last week I talked through AI options with a nonprofit director. The website chatbot came up fast. My answer was no. Keep outreach and the human-touch pages human. AI earns its keep where it makes sense of vast data, not where a visitor expects to reach a person.
That distinction is the whole decision. Chatbots are good at one job: handling the predictable. Hours, pricing, booking, the same five questions the front desk answers every Tuesday. The trend behind the hype is real. AI handled about 30% of customer service cases in 2025, and service teams expect 50% by 2027, per a 2025 Salesforce study cited in Bluehost's 2026 chatbot pricing breakdown. Get the distinction backwards and you pay to annoy your best leads.
Where it actually helps
A trades site is the classic case. Electrician in Missoula, 11pm, homeowner with no hot water. Nobody answers the phone at 11pm, and the caller is not waiting until morning. A chatbot that books the callback beats a contact form every time. That is a chatbot doing a job nobody else can do at that hour.
The bad version is the one you have seen: the bot that asks three clarifying questions, loops you back to the FAQ page, and finally offers to connect you to a human who left at five. Forbes put it bluntly this spring: "Customers Hate Your AI Chatbot. Small Businesses Should Listen." The hate is not for AI. It is for a gatekeeper that never opens the gate.
What it costs to run one
For a small business, expect $20 to $100 a month for a SaaS chatbot. ChatBot.com starts at $19 per user per month billed annually. Intercom starts seats at $29 and bills its Fin agent at $0.99 per resolved outcome. Custom builds start around $10,000 and run past $150,000. Those numbers are from Bluehost's 2026 pricing breakdown.
The part nobody quotes you
The subscription is the cheap part. The expensive part is writing the answers. A chatbot trained on your marketing copy invents confident nonsense. One trained on the actual questions your front desk hears, with the actual answers, is useful on day one. Budget a full afternoon for that. It matters more than which platform you pick. Two failed attempts, then a human. That rule, written into the bot, is what separates the helpful ones from the hateable ones.
Where owners get burned
Chatbot billing has quietly shifted under the old seat model. Platforms now charge per conversation, per AI interaction, or per resolved outcome on top of the seats. A plan that looked like $29 a month turns into $200 when traffic spikes, and seasonal businesses feel it first. Before you sign, get the overage math in writing: what counts as a conversation, what happens when you blow past the included allowance, and whether the AI usage is metered separately from the seat. The vendors who answer those questions plainly are the ones worth buying from.
The one-question test
Before you buy anything, answer one question: is the person on my site trying to reach a human? If yes, skip the chatbot and make the contact path impossible to miss. If no, if they want hours, a price, or a slot on the calendar, a chatbot is a fine employee. It works nights, never calls in sick, and costs less than the coffee budget. The cheapest chatbot on the market is a phone number in the header that somebody actually answers.