--- title: "Website AI Assistant: How to Answer Customer Queries Without a Generic Chatbot" description: "An AI concierge answers only from your pages. See the fail-closed query contract, five escalate triggers, and how unanswered questions become briefs." url: "https://articles.flowcrews.com/website-ai-assistant-how-to-answer-customer-queries-without-a-generic-chatbot" category: "Concierge" date_published: 2026-09-29 reading_time_minutes: 10 word_count: 2367 --- # Website AI Assistant: How to Answer Customer Queries Without a Generic Chatbot *Answer from a cited page, escalate on five triggers, and turn every gap into the next brief.* ![Website AI Assistant: How to Answer Customer Queries Without a Generic Chatbot](https://pub-07fb5e4955ba485b822d6b388be96d9a.r2.dev/2e474d51-a072-472e-ab2f-1a43a8990566/website-ai-assistant-how-to-answer-customer-queries-without-a-generic-chatbot/hero-e3304162-2f84-4866-bb7f-709bf535f5ab.jpg) **TL;DR:** - A useful on-site assistant answers only from your pages and says when it does not know. - Company chatbots are losing the comparison: customers are about three times more likely to use third-party generative AI. - Route tasks; do not invent prices, agreements, or advice the business would have to stand behind. - Escalate on five triggers, not confidence alone, and pass the transcript so the customer does not repeat themselves. - Unanswered questions are the content loop: review them on a set cadence and publish a page the assistant can cite. ## Why a Generic Website Chatbot Loses the Query ## Answer Only From Pages You Control ## Route the Next Step, and Keep Liability With a Person ## Escalate on Five Triggers, Not Confidence Alone ## Turn Every Unanswered Question Into the Next Brief ## Conclusion - Owned-page contract — A website AI concierge is useful only when every allowed sentence comes from a published page the company controls, names that URL, and fails closed instead of guessing. - Published next steps — The assistant may route only documented actions such as tracking an order or opening a page, and prices, agreements, and advice stay with a person. - Label at first contact — From 2 August 2026, EU AI Act Article 50 requires an AI disclosure no later than first contact, so the label is the first message in every market. - Five escalate triggers — Escalate on a person request, no source or low confidence, frustration, sensitive topics, and approval-needed actions, and hand off the transcript, what was tried, and why. - Human share — Plan to staff a 20–30% human share of interactions; an 80% agentic-resolution forecast is a large-center target, not a lean-team one. - Unanswered questions as briefs — Review escalations weekly, cluster by intent, and publish a page only through the existing brief gate before adding its live URL to the allowed source set. Ship the page-only answer contract on your site, then let each unanswered question become the next brief a person approves. ## Frequently Asked Questions ### What is the difference between an AI concierge and a generic chatbot? A generic chatbot fills gaps from general model knowledge. An AI concierge answers from the company’s own pages, cites the page, and says so when it does not know. ### Why are company chatbots losing customers to third-party AI? Gartner’s 2026 survey of 3,566 customers found people about three times more likely to use third-party generative AI than a company chatbot for service issues. Company chatbot use has stayed statistically unchanged since 2022. ### Should the assistant complete tasks, or only answer questions? Among customers who use generative AI, 58% have used it to complete a task, rising to 74% in B2B. The assistant should route actions it is allowed to start, not invent ones that need a person’s approval. ### Who is liable if the assistant gives a wrong answer? The business is. A tribunal in the Air Canada chatbot case rejected the claim that the bot was responsible for its own answers and ruled for the customer. Keep prices, agreements, and advice with a person. ### When must we tell someone they are talking to AI? Label the assistant as AI from the first message. From 2 August 2026, EU AI Act Article 50 requires people in the EU to be told they are talking to an AI, no later than first contact. ### What should happen when the assistant cannot answer? Escalate on an explicit request for a person, with no confirmation loop and no retry, and review the unanswered log on a weekly cadence so gaps become briefs instead of a failure score. Plan for a meaningful share of interactions — at least 20–30% — to still need a human. ## Sources - [https://www.gartner.com/en/newsroom/press-releases/2026-07-08-gartner-survey-finds-customers-are-three-times-more-likely-to-use-third-party-genai-than-company-provided-chatbots-for-customer-service](https://www.gartner.com/en/newsroom/press-releases/2026-07-08-gartner-survey-finds-customers-are-three-times-more-likely-to-use-third-party-genai-than-company-provided-chatbots-for-customer-service) - [https://mementor.no/en/blog/ai/ai-agent-vs-chatbot/](https://mementor.no/en/blog/ai/ai-agent-vs-chatbot/) - [https://www.eesel.ai/blog/how-to-handle-support-escalations-with-ai](https://www.eesel.ai/blog/how-to-handle-support-escalations-with-ai) - [https://www.twig.so/blog/what-happens-when-ai-cannot-answer-customer-question](https://www.twig.so/blog/what-happens-when-ai-cannot-answer-customer-question)