Why hire an external chatbot when your CMS already has one?

Artificial intelligence has now become a must-have for many organisations. Yet all too often, AI projects end up exactly where they don’t belong: with external chatbot services that are tacked onto a company’s own website via a scavenger-hunt-style widget. A second tool, a second contract, a second data flow. And the most important question remains unanswered: why, exactly? Your content management system can already do this itself – if it’s EGOCMS.

The problem with the external chatbot widget

The market for chatbot-as-a-service is vast, and the promises are tempting: ‘Integrated in 5 minutes".’ Many know from experience what actually follows:

• Yet another data silo. The external service has its own knowledge base, its own indexing, its own update logic. Maintaining two systems, keeping two systems in sync.

• Data protection red tape. Every external service requires its own data processing agreements, its own TOM audits, its own data deletion policies. With US providers, the issue of data transfers to third countries is an additional complication.

• Additional costs. Usually billed per conversation or per quota. As traffic grows, so does the bill. Often unexpectedly.

• No rights or client logic. The external chatbot is unaware of your editorial permissions. It does not recognise your clients. It only knows what you can laboriously feed it via API or crawler.

• Discontinuity in the user experience. A third-party widget that does not fit in with the corporate design, either visually or functionally.

In other words: you’re creating a problem to solve a problem that your own CMS has already solved.

The EGOCMS chatbot: native, built into the system, integrated with your content

Since version 25, EGOCMS has taken care of precisely these workarounds. The chatbot is integrated directly into the system as a standard module, not as an external ‘piggyback’ solution. Specifically, this means:

It has access to the entire content repository. The chatbot enables you to query a client’s complete content repository. This is based on the existing Elasticsearch search, combined with semantic search and vector search. The chatbot is therefore more than just a gimmick: it is an alternative, or even a replacement, for traditional keyword search, as it allows search queries in natural language (NLP) and turns them into a genuine conversation.

It respects your access rights. What a visitor is permitted to see in the chat is governed by the same access rights as everywhere else in EGOCMS. An extranet chat responds only with extranet content. An external service is structurally incapable of achieving this.

It is multi-client capable. Multiple categories allow conversations to be restricted to specific areas of knowledge – per client, per target group, per department.

It can be integrated as a page type. Using the dedicated ‘Chatbot’ page type, editors can insert the assistant exactly where it is needed, in line with the corporate design and without writing a single line of code.

AI that won’t cause the data protection officer any sleepless nights

The key point for many of our clients in public administration, higher education and healthcare: the chatbot does not need to be linked to a US cloud. EGOTEC AG can handle the hosting, maintenance and support of open-source LLMs such as Llama or Gemma, hosted in Germany. If you wish, you can also integrate ChatGPT models – or choose not to, opting instead for self-hosting with our lalamo.cloud project.

This ensures that data flows remain where they are already controlled: within your own system, in your own tenant, and under your own TOM. This is precisely what makes the solution attractive for security-critical projects, and it is precisely why, for example, Friedrich Schiller University in Jena uses AI integration in its EGOCMS based on self-hosted LLMs: all data processing takes place within the university network, with fully GDPR-compliant control over data flows and model parameters.

The sustainability and cost-effectiveness are also right: instead of paying a third-party cloud provider for every query, an open-source model runs on the university’s own hardware, conserving resources. Smaller models, used in a more targeted way, are cheaper, more privacy-friendly and, in terms of effectiveness, often barely distinguishable from the giants.

The AI features go far beyond the chatbot

The chatbot is the most visible part of the EGOCMS AI module, but by no means the only one. The module is deeply integrated with EGOCMS and directly supports the editorial team in their day-to-day work:

• AI support for content creation: create, improve, summarise or expand texts directly in the live editor. Simply highlight the text and let the AI make the adjustments.

• AI-powered alternative text: The ‘Alternative Text’ field for images can be populated automatically, including optional AI image recognition. A real boon for accessibility, which has long been mandatory in the public sector.

• DeepL / DeepL Pro: Automatic translation upon saving and publishing; the Pro version includes a custom glossary and workflow. Content can be translated into other languages without the editorial team having to lift a finger.

• Granular deactivation: Individual AI functions can be disabled globally or on a per-client basis. If you do not want AI in a particular area, you can simply switch it off there. This is crucial for operators concerned about data protection.

• Reasoning Models: The chatbot also supports models with reasoning capabilities.

Conclusion

Outsourcing is a good idea if your own system cannot handle the task. With EGOCMS, this is simply not the case when it comes to chatbots. Anyone using a CMS that natively supports AI and complies with data protection regulations is creating unnecessary work, unnecessary costs and unnecessary data protection issues for themselves by using an external chatbot service. The chatbot belongs within the CMS – where the content resides, where the rights lie and where the data is controlled.

Would you like to see how this works? Ask our own chatbot at egocms.com or chatbot.egocms.com – it answers questions about EGOCMS, client management and accessibility round the clock, and, of course, it runs on the very technology it promotes.