Ask for company knowledge instead of searching for it.
Manuals, file shares, project documentation, websites: a knowledge chat based on webx cortex answers questions from your own documents. With source references, and only using the content the person asking is allowed to see.
- Answers with source references
- Permissions applied at search time
- Roles from Active Directory
- Cloud (EU) or local
How an answer is created
Your knowledge stays where it is.
We do not build a new repository. The knowledge chat reads from the sources your team already maintains, and keeps itself up to date when they change.
File shares
Shares on the file server with PDF and text documents. New and changed files are re-read regularly.
Websites and feeds
Your own website, sitemaps and RSS feeds, for example for product information, help pages or news.
Specialist systems
The AI retrieves figures from merchandise management or production through fixed, read-only queries, not through free database access.
We check further formats and systems in the AI potential workshop.
Open to your team, closed to everyone else.
Permissions before the search
Every knowledge source is assigned to roles. The search filters by the roles of the person asking before it returns any results. No approval, no access.
Roles from your directory
On request, users and groups come from your Active Directory via LDAPS. Anyone who leaves the company loses access here as well.
Source references
Every answer names the documents it is based on. This lets your team check every statement directly at the source.
Blocked topics
Terms such as wages, revenue or cost calculation can be blocked. The chat does not answer questions about them.
"Local only" privacy level
Sensitive areas run only with local models. If none is available, there is no answer. More about local AI
Sharing within the team
Good answers can be shared, but only with colleagues who have at least the same permissions. Every share is logged.
Start with one source, then expand.
AI potential workshop
Which questions cost your team time today, where is the answer, who is allowed to see what? The result is a clear first knowledge source with a permissions concept.
Pilot
The knowledge chat runs with this one source and real everyday questions. We measure how often the answers are right.
Expansion
More sources and departments are added. The platform is already in place, so every expansion is faster than the first step.
More from webx cortex.
Local AI
Language models on a server at your premises, for data that should not go into the cloud.
AI agents
From answering to getting things done: agents with fixed tools and approval steps.
AI funding Tyrol
Up to 30% subsidy if the application is submitted before the project starts.
Frequently asked questions
What is AI knowledge management?
AI knowledge management means that employees ask questions in plain language and an AI compiles the answer from the company's own documents, instead of someone searching through folders. Technically, this is usually done with RAG: the documents are captured in a vector database, the relevant passages are searched for each question and passed to the language model as a basis.
What does RAG mean?
RAG stands for retrieval-augmented generation. The language model does not answer from its general training, but on the basis of the text passages previously found in your documents. That is why the answer can name its sources.
Can the AI still say something wrong?
That cannot be ruled out completely with any language model. The risk drops considerably because the AI is restricted to your documents, and every answer names its sources so that your team can check it. In the pilot, we measure how often the answers are right.
Who is allowed to see what?
Every knowledge source is assigned to roles, which on request come from your Active Directory groups. The search takes these roles into account before it returns any results. Content without approval is not found in the first place.
Where is our data stored?
We run the knowledge database on a server in our data centre in Innsbruck or, on request, at your premises. For the answers, webx cortex uses cloud models via EU endpoints or local models, depending on the privacy level. Personal data is masked before every model call.
Which question does your team ask most often?
Tell us where knowledge gets lost at your company. We will tell you honestly whether a knowledge chat helps there. The initial consultation is free of charge.
or by email to info@web-crossing.com · To the AI overview