Services · Company knowledge with AI
Company knowledge with AI: answers your team can check.
Bring approved documents and notes together, let AI answer questions with a link to the source and keep people responsible for checking the result.
Start with ten real questions from daily work and two recurring outputs, and test whether the answers hold up. I set this up with MindKeep or, where integration into existing IT matters, with the Private AI Knowledge Hub that EDIH.li offers with CANCOM.
For whom
- Teams whose answers sit in folders, inboxes and a few people's heads.
- Organisations where new colleagues keep asking the same questions.
- Regulated or trust-sensitive work in finance, construction, legal services and compliance, where results must be traceable.
Not the right fit
- Replacing professional review: answers can be wrong.
- Connecting every system at once: not every drive, office or CRM system has a ready-made connector.
How a knowledge pilot works
01
Choose sources and permissions
Decide which documents and notes are approved and who may see what.
02
Agree real questions
Ten questions from daily work and two recurring outputs, such as a briefing or a standard answer.
03
Answer with sources
The AI answers with references; the team opens the source and checks content and currency.
04
Review and decide
Compare time and rework before and after, test permissions with negative tests and decide whether to expand.
Options
- MindKeep: a shared AI knowledge base for teams with sources, knowledge pages, approvals and roles. In daily use at MMIND.ai and in a pilot. Data in Germany, AI models from Mistral AI. Starts with a free demo.
- Private AI Knowledge Hub (CANCOM, via EDIH.li): a protected AI assistant that finds internal knowledge and prepares answers with sources, matched to your access permissions.
- As a pilot within the 90-day AI transformation: «internal knowledge with references» is one of the two typical pilot cases.
Limits
- Answer quality depends on the content and the model; people check before results are reused.
- Hosting or AI models other than the standard set-up apply only where that configuration has been agreed and checked.
Read more
Questions and answers
How can staff use AI to find the right answer in internal documents?
Approved documents are collected in a protected knowledge base. Staff ask in their own words, and the AI answers with a reference to the source, which they open and check. Permissions decide who sees what.
How is this different from a general chatbot?
Answers come only from the sources your team has approved, each answer links to its source, and checked knowledge is kept apart from raw files. A general chatbot answers from its training and the open web.
Where is the data stored?
With MindKeep, data is stored in Germany in the standard set-up and the AI models come from Mistral AI. For the Private AI Knowledge Hub, storage and permissions follow the agreed CANCOM configuration.
How do we start?
With ten real questions and two recurring outputs, checked together with the team. A free demo or a 30-minute conversation helps choose the first use case.
What does it cost?
The MindKeep demo is free; scope, terms and any paid pilot are agreed separately. Within the 90-day AI transformation, a knowledge pilot can be one of the two pilots included in the monthly price.
Booking opens the EDIH.li business conversation with Hartmut Hübner via Google Meet. Terms, funding and further support are clarified before any work starts.
Useful AI starts with real work.
Bring a concrete question, paper, workflow, or team challenge. We can turn it into a practical next step.