Website Chatbot Basics for Small Businesses

What does a website chatbot do?

A useful starting distinction is between a planned conversation and an AI-generated reply. HubSpot’s rule-based chatbot documentation describes a website widget that sends configured questions and responses. It can gather initial information, qualify leads, book meetings or create support tickets before a team member takes over. Some features require additional subscriptions.

Its AI customer agent instead uses selected business content to answer questions. HubSpot says it can respond with a supporting source, ask for clarification or reassign the conversation to a person, depending on its confidence. These are documented capabilities, not results we have tested.

We use HubSpot here as a documented example, not a comparison or a recommendation to buy it. Our suggested first task for a small business is narrow: help visitors find an approved answer about opening hours, service areas or how to request an appointment.

What should you keep outside its job?

A booking request and a confirmed appointment should be treated as different outcomes. HubSpot describes actions such as booking meetings as configured capabilities. Do not assume a chat widget is connected to your calendar, stock records or customer accounts simply because it can discuss them. Verify each connection and action before describing it as available to visitors.

We recommend keeping exceptions with a person: a disputed refund, an unusual quote, a complaint or a question outside your approved information. The bot should offer a handoff rather than be given authority to invent a policy or promise an exception.

HubSpot’s customer agent overview describes asking visitors to rephrase or transferring to a person when the agent does not know an answer. It also offers reply recommendations that staff can edit, send or dismiss without deploying the agent to live channels. Treat escalation and human review as parts of the design, not evidence that every answer will be correct.

Who owns the answers?

Our recommendation is to name one business owner for the content and one person responsible for incoming handoffs. In a small team, these can be the same person. A vendor or installer can configure the software; your business should approve what it says about your services.

• The answer owner: approves hours, service descriptions, policies and booking instructions; updates them when the business changes; and removes conflicting or expired answers.
• The handoff owner: checks the inbox, handles exceptions and sets an honest expectation for when a visitor will hear back.
• The setup owner: checks which sources the bot can read, which actions it may perform and which staff can change its settings.

This is an operating checklist, not a platform requirement. It follows the work HubSpot documents: selecting content, configuring guidelines and handoffs, testing responses, and reviewing knowledge gaps over time. Keep a short answer register with the approved wording, its source, its owner and its next review date.

Privacy basics before collecting messages

Decide what you need to collect before writing the first question. Our suggested starting point is the visitor’s question and, only when a reply needs it, a contact method. Avoid asking visitors to paste passwords, payment-card details or sensitive records into a general enquiry chat.

HubSpot’s rule-based chatbot guide documents separate settings for consent to collect chat cookies and consent to process data. It lets businesses edit the processing explanation and link their privacy policy. These controls are not a blanket compliance guarantee: HubSpot itself directs businesses to their legal team for advice specific to their situation.

Before launch, review the vendor’s current data documentation and your account settings. Write down which information the bot can access, who can read conversations, how long records are kept, how deletion works and whether data is used for AI training. HubSpot’s customer agent setup guide specifically points to AI settings for controlling shared data and to its AI trust resources. Check the terms for the feature you plan to use; do not infer them from the presence of a consent checkbox.

Make the visitor-facing explanation plain: identify the automated assistant, explain why you request contact details, link your privacy information and offer a way to reach the business. These are our setup recommendations, not a complete legal checklist.

A small launch checklist

HubSpot recommends testing the customer agent’s answers and actions before deployment. We suggest using the following cases with made-up visitor details:

1. Ask a question covered by your approved information. Check the answer and any linked source.
2. Ask about a service you do not offer. Check that the response stays within the agreed scope.
3. Request a person and try again outside business hours. Check who receives the handoff and what the visitor is told.
4. Try a booking request. Check the actual calendar outcome before treating the conversation as a confirmed appointment.
5. Review the privacy explanation and contact collection on a phone as well as a desktop.

Keep the launch scope small enough for the answer owner to review. Record unanswered questions and correct the underlying content before expanding the bot’s responsibilities. This guide makes no claim about lead volume, sales or time saved.