Chatbot Agency for Website and Support

A chatbot agency builds dialogue systems that answer questions on your website or in internal tools, capture requests and route them to the right place. We handle the engineering behind it: knowledge base, answer boundaries, handover to a human, integration with the systems you already run. The bot operates on our own infrastructure in the EU, not inside a third-party SaaS account.

Where chatbot projects fail

The hard part is rarely the language model. It is deciding which content the bot may use and what happens when it does not know something. Without a clean knowledge base it invents answers. Without an escalation path it loses the request as soon as the case gets complicated.

So we settle both before the first line of code: sources, topic boundaries, handover to a real person, logging. And we start small. A bot that reliably answers the five questions your customers ask most and forwards everything else is worth more in daily operation than a general purpose assistant that misses every third answer.

Disclosure under EU AI Act Article 50

Article 50 of the EU AI Act applies from 2 August 2026. Anyone interacting with an AI system has to be able to recognise it. We treat that as part of the build, not as an add-on. Every chatbot we ship states what it is before the first exchange, does not pose as a human and offers a route to a real person. It costs two sentences in the interface and saves you a retrofit under time pressure.

Chatbot, voice agent or agent integration

This page covers the text channel: chat on the website, in a customer portal, in internal tools. When the conversation runs over the phone it belongs to the AI voice agent, because spoken language brings its own requirements for latency, interruptions and recording. When there is no dialogue at all, for example reporting, data preparation or campaign control, our agent integration work is the better frame. Definitions and comparisons, such as the difference between a chatbot and an AI agent, live in our glossary and knowledge base. This page is written for the point where you already want one built.

Technology and operation

Orchestration runs on n8n on our own servers. Conversation history and knowledge base sit in PostgreSQL, editorial content is maintained by your team in NocoDB, so a wording change does not require a deployment. Language models are called through a routing layer, which keeps the bot alive when one provider goes down. Content for the knowledge base comes from your own sources, not from a general web index.

What ships with the bot: a log of all conversations with defined retention periods, handover to email, a ticket system or the phone once the bot reaches its limit, and a review of unanswered questions as the basis for the next iteration.

Working together

We work out of Vienna with small and mid-sized companies in Austria and the wider DACH region. One contact person, direct coordination, no handing the project to a third party.

To get started, a conversation about the use case is enough. Bring the three questions your team answers most often and the systems those answers currently live in. We will tell you whether a chatbot is the right tool or whether a different automation gets there faster. Request a first call.