An AI that knows your company – with your data, not the internet.
ChatGPT knows a lot about the world but nothing about your company. With RAG (retrieval-augmented generation) an AI answers questions based on your own documents – manuals, price lists, contracts, knowledge articles – and shows where each answer comes from.
Was enthalten ist
What RAG is – in short
Your documents are split into sections and made searchable by meaning. When someone asks a question, the system finds the matching passages and the AI answers on that basis only. The result: current, verifiable answers instead of invented ones.
Building the knowledge base
PDFs, Word, SharePoint, Google Drive, Confluence, ticket systems or your website are connected and updated automatically when they change. Permissions are kept – not everyone sees everything.
Internal AI assistant
A chat for your team that answers questions about processes, products and policies. New staff get up to speed faster, experienced staff answer the same question less often.
AI in your software
Summaries, writing suggestions, semantic search, classification or data analysis directly in your website, shop or business application – through clean interfaces instead of copy and paste into ChatGPT.
Measuring quality
Before going live it is tested with real questions: is the answer right, is the source right, does the system know when it does not know? These tests keep running afterwards so changes do not make things worse.
Data protection and data sovereignty
Depending on sensitivity: models hosted in the EU, European providers or open-source models on your own infrastructure. Your data is not used to train third-party models.
Passt zu Ihnen, wenn
- Knowledge sits in folders, PDFs and people’s heads
- Your team keeps searching for the same information
- Your software should get AI features
Wann es nicht das Richtige ist
If your documents are a handful of pages that rarely change, you do not need a RAG architecture – a well-made FAQ or a simple assistant will do. I will tell you what is appropriate.
Ehrlichkeit vorab spart beiden Seiten Zeit — und Ihnen Geld.
Womit gearbeitet wird — und was Sie davon haben
Finds passages that match in meaning, even when the question uses different words than the document.
OpenAI, Anthropic Claude, Mistral or open source such as Llama – chosen for quality, cost and data protection, swappable later.
Answer quality is measured before and after go-live, not judged by gut feeling.
Maintenance - with LivaLink included
Optional and cancellable monthly. Every one of these packages includes LivaLink Business - the subscription alone costs 29 € a month. From 20 € you therefore pay less for maintenance and appointment booking together than for the booking alone.
See LivaLink — appointment booking, contact routes and chat ↗Basic
When you maintain content yourself and only want the technology covered.
- LivaLink Business included - appointment booking, contact routes and chat with your own branding instead of a LivaLink notice (normally 29 € a month)
- Security updates for the system and its extensions
- Daily backup, restore if the worst happens
- Uptime checks - I usually notice an outage before you do
Care
When you would rather hand changes over than make them yourself.
- Everything from Basic
- Content changes: text, images, prices, opening hours - you write me what should change
- Rebuilds of the site structure: new pages, reordering, removing
- Smaller changes are included; anything larger we agree on beforehand
- Reply within one working day
Visibility
When the site should not just run, but keep moving forward.
- Everything from Care
- Ongoing search work: watching rankings, sharpening copy, taking new terms
- Keeping the Google Business Profile and directory listings current
- Building reviews, guided
- A short monthly report: what you are found for and what is next
Without a maintenance agreement your site of course keeps running - you then handle updates yourself, and the regular LivaLink tariff applies.
Häufige Fragen
What is RAG?
Retrieval-augmented generation: before answering, the AI looks up the relevant passages in your own documents and answers on that basis – with sources. Answers stay current and verifiable without retraining a model.
Why not just use ChatGPT?
ChatGPT does not know your internal documents and often may not receive them for data protection reasons. Your own integration accesses your data in a controlled way, respects permissions and fits into your processes.
Does the AI make up answers?
RAG lowers the risk considerably: the AI only answers from sources it found and says when it finds nothing. Tests before go-live check that this works reliably.
Do I need to train my own AI model?
In almost all cases, no. RAG is cheaper, faster to build and easier to keep current than custom training. Fine-tuning only pays off in special cases.
What are the running costs?
Besides hosting there are usage costs for the language model, billed by volume. I calculate them up front for your volume so there are no surprises.
AI integration & RAG für Ihren Betrieb?
Beschreiben Sie kurz, worum es geht. Sie bekommen eine ehrliche Einschätzung zu Aufwand, Dauer und Kosten — und einen Hinweis, wenn etwas anderes besser passt.
No obligation · Reply the same working day