An AI chatbot for business can answer customers at any hour, but what it says, your business has said. This lesson covers how to set one up so it helps customers rather than trapping them, and the rules on telling them it is AI.
In this lesson
- The main kinds of customer-facing bot, and where they help or annoy.
- What a bot needs from you: your own information, and clear rules for handing over to a person.
- How to test one before it goes live.
- Telling customers they are talking to AI, and the rules that apply in the US, the EU and the UK (as of October 2026; not legal advice).
- How to measure whether it is working, and what can go wrong.
Kinds of AI chatbot for business
- Website chat: answers questions, checks order status, books appointments.
- Messaging apps: the same inside WhatsApp, Messenger, Instagram or text messages.
- Phone voice agents: answer calls in a natural voice, take messages, book and route calls.
- Email triage: tags and routes incoming email, drafts replies or sends simple ones.
All of these are agents in the sense of Lesson 7: they act on their own, in front of your customers. They work well for frequent, factual questions answered in your own information (hours, prices, “where is my order?”, bookings), at any hour and without a queue. They annoy customers when they block the way to a person, loop, or meet something emotional with a cheerful script. The fix is always the same: a quick, obvious route to a person.
What a bot needs from you
Your own information
A bot is only as good as what you give it. Before you switch one on, write down:
- prices, services, opening hours and locations;
- your policies (returns, refunds, cancellations, delivery), in plain words;
- the twenty or thirty questions customers ask most, with the answers you want given;
- what the bot must never do: promise refunds, discounts or exceptions, give medical, legal or financial advice, or guess.
Keep this as one dated document and update it when anything changes. Most bot mistakes come from missing or out-of-date information, not from the AI itself.
Clear rules for handing over
The hand-over is the most important part of the design. Decide in advance when the bot passes the conversation to a person, and make sure the person gets the conversation so far, so the customer does not have to repeat themselves.
A customer asks a question. The bot either answers it from your own information (prices, hours, policies), or, if it is unsure, the customer complains, or the customer asks for a person, it passes the customer to a named person with the conversation so far.
Good triggers for a hand-over:
- the customer asks for a person (always honor this, straight away);
- the bot is not sure of the answer, or the answer is not in your information;
- the customer is upset, or mentions a complaint, a safety issue, pain, money lost or a deadline;
- the request is outside what the bot is set up to do (a refund over a limit, an exception to a policy);
- the same question has gone round twice.
Name the person or team who picks up hand-overs, and when. Out of hours, the bot should say when a person will reply, and then someone must.
A business is responsible for what its bot says
In February 2024, a Canadian tribunal ruled on a case that every business with a chatbot should know. A customer asked Air Canada’s website chatbot about bereavement fares, the reduced fares some airlines offer when a family member dies. The chatbot told him he could apply for the reduced rate after travelling, “within 90 days of the date your ticket was issued.” When he did, he “later learned from Air Canada employees that Air Canada did not permit retroactive applications” (Moffatt v. Air Canada, 2024 BCCRT 149).
Air Canada argued it could not be held liable for what the chatbot said. The tribunal member wrote: “In effect, Air Canada suggests the chatbot is a separate legal entity that is responsible for its own actions. This is a remarkable submission.” And: “It should be obvious to Air Canada that it is responsible for all the information on its website. It makes no difference whether the information comes from a static page or a chatbot.”
The tribunal found negligent misrepresentation and ordered Air Canada to pay a total of CA$812.02, of which CA$650.88 was damages and the rest interest and fees. The sum was small; the lesson was not: what your bot tells a customer, your business has told them.
Tell customers it is AI
Customers should know when they are talking to a machine. It is the honest thing to do, it sets expectations, and in a growing number of places it is the law. A simple opening line works: “Hi, I’m the [business] assistant, an AI. I can help with bookings and order questions, or put you through to a person at any time.”
What the rules say, as read on 6 October 2026. This is a summary, not legal advice. Laws change and depend on where your customers are; check with a lawyer or your trade body.
- European Union. Under Article 50 of the EU AI Act, providers must ensure that AI systems “intended to interact directly with natural persons” are designed so that people “are informed that they are interacting with an AI system, unless this is obvious” (AI Act, Article 50). The European Commission says the Act “became applicable on 2 August 2026” and that “chatbots and other interactive AI systems will have to tell users they are dealing with AI, not a human” (Commission; press release, 31 July 2026). It applies if you serve customers in the EU, wherever you are based.
- California. Since July 1, 2019, it has been unlawful to use a bot to communicate with a person in California online “with the intent to mislead the other person about its artificial identity” in order to encourage a sale or influence a vote. A business is not liable under that section “if the person discloses that it is a bot” (Business and Professions Code §17941). The law defines a bot as an automated online account whose actions or posts are all, or substantially all, not the result of a person.
- Utah. Since May 7, 2025, a business using generative AI with consumers must disclose that the person is “interacting with generative artificial intelligence and not a human” if the person clearly asks. People in regulated occupations, such as licensed professions, must disclose it up front in higher-risk interactions (Utah SB 226, 2025).
- Maine. Since September 24, 2025, Maine law has prohibited using an AI chatbot in trade with a consumer in a way that may mislead a reasonable consumer “into believing that the consumer is engaging with a human being unless the consumer is notified” (10 M.R.S. §1500-DD). A violation counts as a violation of the Maine Unfair Trade Practices Act.
- US federal. There is no general federal chatbot disclosure law that we found. A December 2025 executive order directs a federal task force to challenge state AI laws it considers inconsistent with national policy (White House, Executive Order 14365), so the state picture may change. Telling customers anyway is the safe course.
- United Kingdom. We found no UK law that specifically requires a chatbot to say it is AI. The government’s AI principles say “AI systems should be appropriately transparent and explainable,” including information on “how, when, and for which purposes an AI system is being used” (AI regulation white paper). If your bot handles personal data, the Information Commissioner’s Office says “you need to be transparent about how you process personal data in an AI system” (ICO); the ICO notes that guidance is under review.
Keep conversation logs for a sensible period, in line with your privacy notice (which should say conversations are recorded), so you can see what the bot told a customer if there is a dispute.
Test it before launch
- Write twenty real questions from your emails, calls and messages, including the awkward ones: the angry customer, the request for an exception, the question your information does not answer, someone trying to talk it into a discount.
- Run them all and mark each answer: right, wrong, or should have handed over.
- Fix the information and the rules, not just the wording, then run them again.
- Have someone outside the business try it, including a person who uses a screen reader. Check the chat works with a keyboard and the phone agent copes with accents and background noise.
- Start small: out of hours only, or one kind of question, with a person reading every conversation for the first weeks. Limit what the bot can see and do, because people will try to trick it (Lesson 10).
Measure it
Three numbers tell you most of what you need:
- Hand-over rate: the share of conversations passed to a person. Too low can mean the bot is answering things it should not; too high means it is not saving much.
- Resolution: the share of conversations where the customer got what they needed without coming back. Ask a one-click question at the end, and check a sample yourself.
- Complaints: complaints that mention the bot, and any wrong answers found in spot checks.
Read a sample of conversations every week. You will learn more from twenty transcripts than from any dashboard. Most problems are made-up policies, out-of-date facts or customers who cannot reach a person, and each traces back to the information sheet or the hand-over rules.
Tools and the roles they touch
Customer-facing bots mostly change the work of receptionists and customer service staff. Our AI tools for receptionists and for customer service pages list tools with prices as the vendors list them. Each job page shows which of the role’s tasks our data rates needs a human: those are the conversations your bot should hand over.
Key terms
- AgentAn AI given a goal, tools and permission to take several steps on its own, such as booking, filling in forms or updating records.Because agents act, their mistakes have effects. Start with a person approving every action.Lesson 7: From chatbot to agent · More in the glossary
- Retrieval (RAG)Giving a model your own documents to work from, so it answers from them rather than from memory. RAG stands for retrieval-augmented generation.More in the glossary
- HallucinationWhen AI produces something that sounds right but is false, such as an invented fact, policy or source.Giving it your documents, asking for sources and checking its work make it much rarer and easier to catch.Why AI makes things up · More in the glossary
- Human in the loopA named person who checks or approves AI’s work before it takes effect, and answers for the result.Lesson 10: Risks and rules · More in the glossary
- Prompt injectionTricking an AI with instructions hidden in what it reads, such as an email or web page, so it does something it should not.Limit what an agent can do, and have a person approve anything it sends, pays or deletes.OWASP: Prompt injection · More in the glossary
Every term is explained in the glossary.
Try this
Collect the last twenty questions customers asked you by phone, email or message. For each, write whether a bot could answer it from information you already have written down, whether it should always go to a person, or whether you would need to write the answer down first. If fewer than half could be answered from written information, start by writing that information: it will help your staff even if you never use a bot.
Sources
- Moffatt v. Air Canada, 2024 BCCRT 149, Civil Resolution Tribunal of British Columbia, issued 14 February 2024. Read 6 October 2026. https://decisions.civilresolutionbc.ca/crt/crtd/en/item/525448/index.do
- Regulation (EU) 2024/1689 (Artificial Intelligence Act), Article 50, consolidated text as at 27 July 2026, AI Act Service Desk. Read 6 October 2026. https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-50
- European Commission, “Regulatory framework on AI”, last updated 3 August 2026. Read 6 October 2026. https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai
- European Commission, “Commission starts enforcing AI Act rules and new transparency requirements”, 31 July 2026. Read 6 October 2026. https://digital-strategy.ec.europa.eu/en/news/commission-starts-enforcing-ai-act-rules-and-new-transparency-requirements-2-august
- California Business and Professions Code §§17940–17943 (SB 1001, 2018), California Legislative Information. Read 6 October 2026 (Internet Archive copy of 28 September 2026). https://leginfo.legislature.ca.gov/faces/codes_displayText.xhtml?lawCode=BPC&division=7.&title=&part=3.&chapter=6.&article=
- Utah Legislature, SB 226 (2025), enrolled copy. Read 6 October 2026. https://le.utah.gov/Session/2025/bills/enrolled/SB0226.pdf
- Maine Revised Statutes, Title 10, §1500-DD, “Required disclosure of use of artificial intelligence chatbot to engage in trade and commerce” (PL 2025, c. 294). Read 6 October 2026. https://legislature.maine.gov/statutes/10/title10sec1500-DD.html
- Maine Legislature, Legislative Digest, 132nd Legislature (LD 1727 is Public Law 294; effective date September 24, 2025). Read 6 October 2026. https://legislature.maine.gov/doc/11927
- The White House, Executive Order 14365, “Ensuring a National Policy Framework for Artificial Intelligence”, 11 December 2025. Read 6 October 2026. https://www.whitehouse.gov/presidential-actions/2025/12/eliminating-state-law-obstruction-of-national-artificial-intelligence-policy/
- UK Government, “A pro-innovation approach to AI regulation” (white paper), 29 March 2023, updated 3 August 2023. Read 6 October 2026. https://www.gov.uk/government/publications/ai-regulation-a-pro-innovation-approach/white-paper
- Information Commissioner’s Office, “How do we ensure transparency in AI?”, 15 March 2023. Read 6 October 2026. https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-resources/artificial-intelligence/guidance-on-ai-and-data-protection/how-do-we-ensure-transparency-in-ai/