# RubyRep — full reference Your website's AI representative. RubyRep learns your website and answers your customers on it — 24/7, in your words, with links to your own pages. Install one line of code. --- ## AI customer service URL: https://rubyrep.com/glossary/ai-customer-service Definition: AI customer service is the use of a language model to answer customer questions automatically, usually on a website, in chat, or over email. AI customer service means answering customer questions with a language model rather than a person. The useful implementations are grounded in a specific business's own content; the poor ones let a general model improvise, which produces confident answers that are wrong. Why it matters: Most customer questions are repetitive and already answered somewhere in a company's published material. Automating those frees a support team for the problems that genuinely need judgement — but only if the system is grounded in real company content rather than a model's general knowledge. ## RAG (retrieval-augmented generation) URL: https://rubyrep.com/glossary/rag Definition: Retrieval-augmented generation (RAG) is a technique where relevant source passages are retrieved from a document collection and given to a language model, so its answer is grounded in those passages rather than its training data. RAG is what separates a chatbot that knows your business from one that guesses. Documents are split into passages and indexed by meaning; each question retrieves the closest passages, and the model is instructed to answer only from them. Why it matters: A language model on its own has no reliable knowledge of your prices, policies or products, and will fabricate plausible answers if asked. Retrieval puts the real text in front of it at the moment of answering, which is both more accurate and auditable — you can show which passage an answer came from. ## Knowledge base URL: https://rubyrep.com/glossary/knowledge-base Definition: A knowledge base is an organised collection of articles and documents that answers recurring questions about a product, service or organisation. Traditionally a knowledge base is a help centre humans read. For AI support it is the source material a representative answers from — which means an ordinary website, well written, can serve as one without a separate help centre. Why it matters: The quality of an AI representative is capped by the quality of its knowledge base. A site that clearly states pricing, policies and product details produces good answers; a site that hides them behind marketing copy produces vague ones. ## Website chatbot URL: https://rubyrep.com/glossary/website-chatbot Definition: A website chatbot is a chat interface embedded on a website that answers visitor questions, either from a scripted decision tree or by generating answers from source content. The term covers two very different things. Older chatbots follow scripts and break the moment a visitor phrases something unexpectedly. Modern ones retrieve from a business's own content and generate an answer, which handles arbitrary phrasing. Why it matters: Visitors judge a chatbot in one message. A scripted bot that answers "I didn't understand that" trains people never to use it again — and to email you instead, which is the outcome you were trying to avoid. ## AI agent URL: https://rubyrep.com/glossary/ai-agent Definition: An AI agent is a system that uses a language model to work towards a goal over multiple steps, typically by deciding which tools or actions to use along the way. The word is used loosely. In customer support it usually means something more capable than a question-answering bot — a system that can look things up, take an action, or decide to escalate, rather than only replying. Why it matters: The distinction matters when buying. A system described as an "agent" may be able to act on your systems, which is powerful and also a much larger security and correctness question than answering from published content. ## Embedding URL: https://rubyrep.com/glossary/embedding Definition: An embedding is a numerical representation of text that captures its meaning, allowing passages with similar meaning to be found even when they share no words. Embeddings are why an AI representative can answer "do you deliver to Berlin?" from a page that only says "we ship across the EU". Text is converted to vectors, and similar meaning lands close together. Why it matters: Keyword search fails on paraphrase, which is how customers actually write. Embeddings make retrieval robust to phrasing, which is the difference between a bot that usually finds the answer and one that usually doesn't. ## Lead qualification URL: https://rubyrep.com/glossary/lead-qualification Definition: Lead qualification is the process of deciding which enquiries are a genuine fit for what you sell, so sales time goes to the ones most likely to convert. Traditionally a form or a discovery call. Increasingly it happens before contact: a visitor who can get straight answers on price range, scope and coverage qualifies themselves, and only books when there's a fit. Why it matters: Qualification that happens after a call has already cost you the call. Publishing the answers — and making them easy to get to — moves the filter earlier, which is cheaper for both sides. ## Answer engine optimization (AEO) URL: https://rubyrep.com/glossary/answer-engine-optimization Definition: Answer engine optimization is the practice of structuring content so that AI-powered answer engines can understand it, quote it accurately, and attribute it to you. Search increasingly returns synthesised answers rather than links. AEO is the unglamorous work of making sure the facts about your business — what it is, who it's for, what it costs — are stated plainly enough to be quoted correctly. Why it matters: If an answer engine can't find a clear statement of your pricing or capabilities, it will either omit you or infer something wrong. Clear, factual, well-structured pages are the defence. ## 24/7 customer support without night shifts URL: https://rubyrep.com/use-cases/24-7-customer-support Round-the-clock cover normally means hiring across time zones or outsourcing to a team that doesn't know your business. An AI representative trained on your own website answers the majority of out-of-hours questions correctly, and hands the rest to you in the morning. Problem: Questions don't stop at 6pm. Most businesses answer them the next working day, by which point the customer has bought elsewhere or lost interest. How it works: - RubyRep reads your website and builds a knowledge base from your existing pages. - The widget answers visitors at any hour, using only what your site says. - When it can't answer, it says so and offers your contact options rather than guessing. - Every out-of-hours conversation is waiting in your dashboard the next morning. ## FAQ automation that uses your real pages URL: https://rubyrep.com/use-cases/faq-automation A static FAQ page ages badly and never quite matches how people phrase things. Reading your whole site instead means the answer to a question can come from wherever it genuinely lives — a policy page, a product spec, a help article. Problem: FAQ pages are written once, drift out of date, and only answer questions in the exact words the author imagined. How it works: - RubyRep reads every page, not just the FAQ, so answers can come from anywhere on your site. - Visitors ask in their own words; retrieval finds the relevant passage regardless of phrasing. - Questions that come up repeatedly are surfaced in your dashboard as candidates to document properly. ## After-hours support that doesn't lose the lead URL: https://rubyrep.com/use-cases/after-hours-support The gap between a question and an answer is where interest dies. Answering after-hours questions immediately — and routing the ones you can't answer to a form or booking link — keeps enquiries warm until someone is back at a desk. Problem: Out-of-hours enquiries either go unanswered until morning or hit a contact form that feels like a void. How it works: - Visitors get a real answer at the moment they ask, whatever the hour. - Where a human is genuinely needed, RubyRep offers your booking link or contact details immediately. - The full conversation is in your dashboard, so whoever picks it up has the context. ## Turn website questions into qualified enquiries URL: https://rubyrep.com/use-cases/website-lead-capture Most visitors who leave without contacting you had one unanswered question. Answering it on the spot — then offering a booking link at the right moment — converts curiosity into a conversation with your team. Problem: Contact forms ask for commitment before the visitor has the information they need to commit. How it works: - RubyRep answers the practical questions that precede an enquiry: scope, price range, process, coverage. - When the visitor is clearly ready, it offers your booking or contact route. - You see the full conversation, so the first human reply isn't starting from zero. ## Answer customer questions from your own content URL: https://rubyrep.com/use-cases/customer-question-answering Answering questions well means answering them from something true. RubyRep retrieves the relevant passages from your site before it writes anything, and cites the page it used, so an answer can always be checked. Problem: General AI assistants sound confident about businesses they know nothing about, which is worse than no answer at all. How it works: - Your pages are split into passages and indexed for meaning, not just keywords. - Each question retrieves the most relevant passages from your site. - The representative answers from those passages only, and links the source page. - If retrieval finds nothing solid, it says it doesn't know. ## Qualify enquiries before they reach your team URL: https://rubyrep.com/use-cases/lead-qualification The fastest qualification is honest information, early. When a visitor can find out your minimum project size or which regions you serve without booking a call, the calls you do take are with people who fit. Problem: Sales calls get spent re-explaining scope, pricing bands and coverage to people who were never a fit. How it works: - RubyRep answers scope, pricing and coverage questions from your published pages. - Visitors who aren't a fit find that out immediately, without taking a slot. - Visitors who are get pointed to your booking link at the right moment. ## AI customer support for SaaS companies URL: https://rubyrep.com/ai-customer-support/saas SaaS support is dominated by questions your documentation already answers: what a plan includes, whether a feature exists, how an integration works. RubyRep reads your docs, pricing and changelog and answers those on the site, so trial users get unblocked in seconds instead of waiting on a ticket. Common visitor questions: - Does the Pro plan include SSO? - How many seats come with the team plan? - Do you have a Zapier integration? - What happens to my data if I cancel? - Is there an API rate limit? ## AI customer support for ecommerce stores URL: https://rubyrep.com/ai-customer-support/ecommerce Most ecommerce support is four questions wearing different hats: where is it, when will it arrive, can I return it, and will it fit. RubyRep reads your product pages and policies and answers them at the moment of hesitation, which is the moment the sale is won or lost. Common visitor questions: - Do you ship to Canada? - How long do returns take to refund? - Is this jacket true to size? - What's the difference between these two models? - Is this back in stock soon? ## AI customer support for real estate URL: https://rubyrep.com/ai-customer-support/real-estate Property enquiries arrive at night and on weekends, and the first agent to reply usually wins the viewing. RubyRep answers questions about listings, fees and the buying or renting process from your own site, then hands warm enquiries to an agent with the contact options you set. Common visitor questions: - What's the deposit on this rental? - Are pets allowed in this building? - What fees do buyers pay? - How do I book a viewing? - Which areas do you cover? ## AI customer support for healthcare practices URL: https://rubyrep.com/ai-customer-support/healthcare Most calls to a practice are logistics, not medicine: opening hours, what a visit costs, which insurance is accepted, what to bring. RubyRep answers those from your website so reception can focus on patients in front of them — and it stays firmly out of clinical advice. Common visitor questions: - Do you take my insurance? - How much is a first appointment? - What should I bring to my visit? - Where do I park? - Do you offer evening appointments? ## AI customer support for schools and course providers URL: https://rubyrep.com/ai-customer-support/education Admissions seasons concentrate a year of questions into a few weeks, and they are the same questions every year: entry requirements, deadlines, fees, term dates. RubyRep answers them from your own prospectus and admissions pages, at whatever hour a parent or applicant is actually researching. Common visitor questions: - What are the entry requirements? - When is the application deadline? - How much are tuition fees? - Do you offer scholarships? - When does the term start? ## AI customer support for agencies URL: https://rubyrep.com/ai-customer-support/agencies Agencies field two kinds of question: what do you do, and what does it cost. Both are answerable from the site, and both currently interrupt someone senior. RubyRep answers them from your services and case study pages, and doubles as something you can deploy on client sites. Common visitor questions: - What's your typical project size? - Do you work with startups? - How long does a rebrand take? - Do you do retainers or project work? - Which industries have you worked in? ## How to add an AI chatbot to your website URL: https://rubyrep.com/guides/how-to-add-ai-chatbot-to-website Adding an AI chatbot takes minutes. Making it answer correctly takes preparation. This covers what to check before you install anything, how installation actually works, and the three failure modes that make businesses take chatbots back down. ### Check your content before you check vendors An AI chatbot answers from source material. If your pricing lives in a PDF, your policies are three sentences long, and your product pages are mostly imagery, every chatbot you try will be vague — because there is nothing specific to be accurate about. Before evaluating tools, read your own site as a stranger. Can you find what a product costs, how returns work, and how to reach a person? If not, fix that first; it improves your conversion rate whether or not you ever install a chatbot. ### Decide what it is allowed to do The single biggest quality decision is scope. A chatbot that answers questions from published content is a well-defined, low-risk system. A chatbot wired into your order system can act on a customer's account, which is a different category of project. Start with the first. Write down explicitly what is out of scope — account-specific queries, medical or legal advice, anything requiring judgement — and make sure the tool you pick will hand those to a human rather than attempt them. ### Install it Modern chatbots install with a single script tag before the closing body tag. The script should load asynchronously, render inside an iframe or shadow root so it cannot collide with your CSS, and add no meaningful weight to your page. - Train the chatbot on your website by giving it your URL. - Review the answers it gives to your five most common questions. - Copy the embed snippet and paste it before in your site template. - Load a page and confirm the launcher appears and answers correctly. - Configure where it hands over to a human — email, booking link, or phone. ### The three ways this goes wrong It invents answers. If a chatbot will not say "I don't know", it will eventually tell a customer something untrue about your refund policy. Test this deliberately: ask something your site does not cover and see what happens. It goes stale. Content changes and the chatbot keeps quoting last quarter's pricing. Whatever you install needs a re-crawl you will actually run, ideally on a schedule. It has no exit. A customer who needs a person and cannot find one is more annoyed than if the chatbot had never existed. The handover route should be one click, always visible. ## How to train AI on your website content URL: https://rubyrep.com/guides/how-to-train-ai-on-website-content "Training on your website" almost never means training a model. It means crawling your pages, splitting them into passages, indexing them by meaning, and retrieving the right ones at question time. Understanding that distinction tells you exactly what to fix when answers are wrong. ### It isn't training, it's retrieval When a product says it trains on your website, it almost always means retrieval-augmented generation: your pages are crawled, split into passages, converted into embeddings, and stored. When a visitor asks something, the closest passages are retrieved and handed to a language model with an instruction to answer only from them. This matters practically. Nothing about the model changes, so updating your knowledge is a re-crawl, not a retraining run — and a wrong answer is usually a retrieval problem or a content problem, not a model problem. ### What makes content retrievable Retrieval works on meaning, so content answers well when it states things plainly and completely in one place. A page that says "we ship across the EU in 5–8 business days, free over €80" is retrievable. A page that spreads the same facts across a hero image, a tooltip and a PDF is not. Length matters in both directions. A page with forty words has nothing to retrieve; a ten-thousand-word page gets split into passages that may each lack the context to stand alone. Pages of a few hundred to a couple of thousand words, each about one thing, work best. ### Diagnosing a wrong answer Work backwards through the chain. First: does the fact exist on the site as text? If not, that is the fix — write it. Second: was the right passage retrieved? If retrieval surfaced the wrong page, the content probably buries the answer or uses vocabulary customers don't. Third: was the passage retrieved but the answer still wrong? That is a prompt problem, and it usually means the system is not being firm enough about answering only from source. ### Keep it current Set a cadence that matches how often you change things. An ecommerce store during a sale needs re-crawling more often than a professional services firm. The failure mode is silent — nothing breaks, the answers just quietly describe an older version of your business. ## How to automate customer support without making it worse URL: https://rubyrep.com/guides/how-to-automate-customer-support Support automation fails when it is applied to the wrong tickets. The reliable approach is to automate the repetitive, publicly-answerable layer completely, route everything else to a person quickly, and measure the handover rate rather than the deflection rate. ### Sort your tickets before you automate anything Pull a month of tickets and sort them into three piles: questions your website already answers, questions that need account access, and questions that need judgement. Only the first pile is safely automatable from published content. Most teams find the first pile is larger than they expected — often a third to a half of volume. That is your realistic ceiling, and it is worth knowing before a vendor quotes you a number. ### Automate the first pile completely Half-automation is the worst outcome: a bot that answers some shipping questions and fumbles others teaches customers not to trust it, and they escalate everything anyway. If a category is automatable, cover it properly — which usually means improving the underlying content, not tuning the bot. ### Make the exit obvious Every automated interaction needs a visible, one-click route to a person. Hiding it to protect deflection numbers is a false economy: it converts a mild annoyance into a complaint, and the customer reaches you anyway, angrier. ### Measure the right thing Deflection rate is easy to game and tells you little. Better measures: how many automated conversations ended without a follow-up contact, how many were rated unhelpful, and what your team's ticket mix looks like a month later. If the repetitive tickets are gone and the hard ones remain, it worked.