omesos
Book an assessment

For companies · Private document assistant

A private AI assistant for your company documents.

In your own AWS cloud, set up and run for you. Your team gets answers from your own documents with the sources shown, and your data protection officer and works council get everything they need to say yes.

What your team can do with it

One place where your staff ask questions in plain language and get answers from your own documents, in German or English. It runs in your cloud, and we keep it running.

  1. Ask across your documents

    Questions in plain language, answered from the collections you have approved. “What does our travel policy say about first-class rail?” gets an answer without anyone digging through folders.

  2. See the sources

    Every answer names the documents and passages it draws on, so people can check it before they rely on it.

  3. Summarise

    Long contracts, reports and meeting minutes condensed to the points that matter, with follow-up questions whenever something needs a closer look.

  4. Draft

    First versions of emails, letters, replies and internal texts, built on your own material and ready for a person to review and finish.

  5. In German and English

    Ask in German or English and get the answer in the same language. Before anything goes live, we test several models on your own documents in both languages and choose the one that answers best.

Model catalogue

The open models that matter, running in your cloud.

Open weights mean the whole model runs inside your cloud. Your prompts never reach the company that trained it, wherever it comes from. We test the candidates on your own examples, check each licence for your use, and keep the catalogue current as new releases appear.

  1. DeepSeekDeepSeek
  2. QwenAlibaba
  3. GLMZhipu AI
  4. KimiMoonshot AI
  5. LlamaMeta
  6. MistralMistral AI
  7. GemmaGoogle
  8. gpt-ossOpenAI
  9. and many more

Model names and logos are trademarks of their respective owners. No affiliation or endorsement is implied.

Which documents it reads

The assistant reads the sources you choose and nothing else. Which ones, we work out together.

Typical collections

  • Policies and handbooks
  • Contracts and templates
  • Project files
  • Product and technical documentation
  • Meeting minutes
  • Internal FAQs

  1. You choose the sources

    You decide which collections the assistant may read. Nothing is connected just because it happens to be reachable.

  2. We check them together

    In the assessment we go through the sources with you: where they are stored, which formats they hold, how current they are, and who should be able to ask about them.

  3. What stays out, stays out

    We agree just as clearly what the assistant must not read, for example personnel files, and put it in writing for your data protection officer and works council.

  4. Connected one by one

    Each source is connected only after you approve it. Start with one collection and add more once it works well for your team.

Where your documents stay

Who can see what, in detail
  • Everything runs in your own cloud, in the EU region you choose, for example Frankfurt.
  • Your prompts and documents are not sent to any AI model vendor.
  • You own the keys and can revoke our access at any time.
  • The model runs on Amazon Bedrock, the managed AWS model service, or on dedicated GPU instances in your cloud, whichever fits your documents, your users and your budget.
  • On the dedicated-GPU setup, the model version stays fixed until you approve a change, and AWS commits in its Service Terms that its staff have no technical means to access your content on Nitro EC2 instances.

Designed for approval

An AI assistant your company provides usually needs the agreement of your works council (co-determination under § 87(1) no. 6 BetrVG), and your data protection officer will want to see the paperwork. We plan for both conversations from day one, not after the pilot.

For your works council

Clear logging, no performance monitoring

No performance monitoring
No usage reports, rankings or productivity dashboards per person. None of these is part of the setup, and none is added without your works council’s agreement.
Technical logs for operation only
Errors, load and overall volume: what it takes to keep the system running and secure. We set up logging so that the text of questions and answers stays out of it.
A retention period you agree on
Chat histories stay in your cloud and are deleted after the period you agree with your works council.
A written logging description
What is recorded, why, for how long, and who may look at it on which occasion. Written so it can go straight into your works agreement.
Questions answered first-hand
We explain the setup to your works council directly and answer the technical questions of any expert it brings in. Where a works council has to assess AI, the law treats bringing in an expert as necessary (§ 80(3) BetrVG).

For your data protection officer

The paperwork, before the first document

Processing agreement
A GDPR processing agreement under Art. 28 GDPR (AV-Vertrag) between you and us.
Secrecy agreement
A separate secrecy agreement on top of the processing agreement.
Input for your DPIA
Material for your data protection impact assessment: what is processed, where, by whom, and with which safeguards.
AI notice for users
A ready text that tells your staff they are working with an AI system and what it is meant for.
Licence register
A list of every model and software component in the setup, with its licence.
Our access, on the record
Every access by us is time-limited and logged in your own CloudTrail.

The processing and secrecy agreements bind us as your operator; your contract with AWS stays directly between you and AWS. As the controller, you keep the GDPR decisions in your own hands, and we prepare the material you need for them.

Rollout in steps

Small steps, and you approve each one before the next begins.

  1. Assessment

    Thirty minutes on the documents you have in mind, the number of users, your clouds and region, how your staff sign in, and what your data protection officer and works council will ask. You leave with a concrete recommendation.

  2. Paperwork first

    The processing agreement, the secrecy agreement, the logging description and the DPIA input are on the table before a single document is connected. Your works council sees the same papers as your data protection officer.

  3. Setup in your cloud

    Chat, document search and the model backend, set up in your cloud and the EU region you choose. We work through an access role you grant and can revoke at any time.

  4. Pilot with a small group

    A group you pick works with a first collection. We test candidate open-weight models, for example from the Llama, Mistral, Qwen or Gemma families, on questions your pilot group writes in German and English, and settle on what answers best within your budget. You judge the results on your own material.

  5. Rollout and operation

    Then the rollout to everyone it is meant for, and ongoing operation for a monthly fee: updates, monitoring, and a named person to talk to during business hours (CET). When a better model appears, we test it on your documents and switch only after your approval.

And if you ever want to stop: revoke our access. The setup, your documents and the logs stay in your cloud.

What it costs

Thirty minutes to see how it would work for you

Bring the documents you have in mind and the first questions from your data protection officer and works council. You leave with a concrete recommendation: which sources to start with, which setup fits, and what the rollout would look like.

Book an assessment