deisenwerk

deisenwerk GmbH

Self-hosted AI automation

We build AI-powered workflows for your inbox, documents and internal knowledge — with self-hosted components, so your data never leaves your premises.

Discuss your project

Our promise

Your data stays with you

AI value without data leakage. Language models, vector databases and processing pipelines run in your cloud or on-premises — not at OpenAI, not at Anthropic, not at Google.

  • Data sovereignty — PII, contracts and internal documents never leave your infrastructure.
  • Compliance-friendly — GDPR, FINMA, attorney-client privilege: we work within the constraints that apply to you.
  • No vendor lock-in — open-source models and tooling, auditable and replaceable.

Use cases

Mailbot — intelligent inbox

Classify incoming email, extract order details, contracts and invoices, draft replies. Your team decides — the AI handles the routine work.

Hours per week recovered, fewer missed requests.

Customer correspondence stays on your server — no uploads to external AI providers.

Typical: Pilot (4 weeks) → Production implementation (3–4 months)

Document processing (PDFs, forms, KYC)

Structured extraction from PDFs, scanned forms and contracts. Validation against internal systems, routed into existing workflows.

Manual data entry replaced by automated pipelines — at equal or higher accuracy.

Sensitive documents (KYC, HR, contracts) are processed on-premises — never sent to third parties.

Typical: Assessment + Pilot (4–6 weeks) → Production implementation (4–6 months)

Internal knowledge assistant (RAG)

Chat interface over your wikis, SharePoint and code repositories. Employees ask in natural language — answers come with source citations.

Onboarding accelerated, expertise accessible, repeat questions reduced.

Your internal content isn't used to train someone else's model — everything stays with you.

Typical: Pilot (3–5 weeks) → Production implementation (3–5 months)

0+Years of experience
0Regulated clients
0ETHGlobal prizes
0Years as CTO

How we work

Four stages, each self-contained — most engagements start at one stage and grow as needed.

  1. 01

    AI readiness assessment

    1–2 weeks

    Structured review of your use cases, data, infrastructure and in-house skills.

    Outcome
    Written recommendation on what is worth building locally, what to handle via API, and what to defer.
    Suited when
    You know AI is relevant to your domain, but you are unsure where the first meaningful step lies.
  2. 02

    Pilot / proof of concept

    3–6 weeks

    One use case end-to-end on your infrastructure — from data connectors to user interface.

    Outcome
    Working artifact, measured outcome, solid basis for the go/no-go decision.
    Suited when
    You have a concrete use case in mind and want to test effort, quality and value before a larger investment.
  3. 03

    Production implementation

    8–16 weeks, scope-dependent

    Full deployment including integration with existing systems, access control, documentation and handover.

    Outcome
    Solution running on your infrastructure with clear ownership and a maintenance path.
    Suited when
    A pilot or equivalent groundwork is in place; you are ready to operate the solution and use it internally.
  4. 04

    Operation & evolution

    ongoing

    Monitoring, model updates, scaling and advisory during day-to-day operation — as a retainer or on demand.

    Outcome
    Systems that stay reliable, planned evolution, no standstill when models or infrastructure change.
    Suited when
    A production solution is in place and you want a long-term partner instead of a new project for every change.

About us

10+ years of engineering — from banking core systems to AI hackathon prizes.

Deep technical and regulatory background, grown in production with regulated clients.

See our background

Contact

Let's discuss your project.

deisenwerk GmbH

Handelsregister-Nr: CH-035.4.036.708-1

UID/MWST: CHE-113.266.074