AI chatbot and AI integration for websites inside the site you already have

In most cases yes. An AI chatbot for a website connects to the existing site through a plugin, an embed script or an API call to our server, so there is no need to replace the platform. On a WordPress or WooCommerce site that is usually a plugin plus a server layer, on Shopify an app or a theme embed, and on Wix an external component embed. On a custom-code site or an app we connect directly. Some platforms limit custom code, and we check that in the first conversation before committing to a particular solution. The right route depends on the site version, the plugins already installed and the capability you want to add.

The assistant gets access only to content you approved: the site pages, the catalogue, terms documents and frequently asked questions. Every answer rests on a passage found in them, and if no suitable passage exists it says it has no information and offers to take contact details or hand over to a person. We add written rules for sensitive subjects, such as discounts and commitments, and test it beforehand on dozens of questions from your business. After launch we read sample conversations and fix the sources when a gap appears. The rules and sources can be updated at any time without touching code, and your team sees which questions were asked and which went unanswered.

We connect OpenAI, Claude and Gemini through their APIs and choose by task: a fast, inexpensive model for classification and search, a stronger one for complex conversations. Each provider publishes its terms on data retention and training for API use, and we review them and show them to you before we start. You can also configure the system so that identifying details are never sent at all. The API keys belong to your own account, so costs and data stay under your control. For each capability we also define which fields may be sent to the provider and which stay with you, and we document it in writing so customers and staff know exactly what happens to their data.

Running cost depends on usage volume, because providers charge by the amount of text processed. To avoid surprises we set a monthly usage ceiling with an alert as it approaches, a message limit per visitor, caching for repeated questions, and the smallest model that is good enough for each task. A dashboard shows what each capability cost and how many conversations it handled, so you can decide whether to expand or stop. Provider prices change, so you receive the current figures in the quotation rather than from a page like this one. Provider costs are also checked against real site traffic, so you know in advance what will happen in a peak season or after a large paid campaign, and the ceiling protects you either way.

Amendment 13 broadened the obligations of businesses that hold personal data, so privacy is part of the specification. In practice we work on data minimisation: anything not needed for an answer is not sent to the model, identifying details in a conversation are removed or replaced where possible, and the visitor sees a notice that they are talking to an automated system. Conversations are stored in an environment you control for a period you set, and access is logged. We do not give legal advice; if you need a privacy adviser to review it, we prepare an exact description of the data flow. Where the site serves a wide audience we also add a link to the privacy policy and a way to leave the conversation and ask for deletion of whatever was stored about it.

AI integration for websites is the customer-facing implementation: what the visitor sees, where they meet the capability and how it fits into browsing, the catalogue and the admin. Language model and RAG integration deals with the knowledge and document layer behind the answers, and an AI agent carries out actions in your business systems. In a single project we often combine several of them. If all you need is a chat over internal documents for staff, or a WhatsApp agent with no website, we point you to the matching service and do not sell more than is needed. The choice depends on the business goal: what matters most to you on the site, and which of your existing systems should receive the result of each conversation.