DE EN

AI engineering and process delivery. For mid-sized companies and enterprises.

Let’s take a look together
Our core offer

Managed AI Automation

We automate business processes that are still done by hand, over and over. With AI, we can also automate the steps that used to need a person.

Find

Two to three days on site. Afterwards, the first workflow is defined, with a start, an end and clear responsibilities.

Build

Standard case, exceptions and acceptance criteria defined, then tested with real cases. First production version usually within two to four weeks.

Run

We monitor the live workflow, handle incidents and agree on changes with you.

Fixed rules handle the clear-cut steps. AI makes sense of the content. Your team decides where responsibility remains.

Our goal for the first workflow: take at least four hours of recurring work per week off your team.

Safety nets for every automation
  • Dry run · logs first, sends nothing
  • Frequency cap · limited messages per contact
  • Send window · business hours only
  • Reply detected · automation pauses, task goes to the owner
  • Kill switch · everything stops, log per contact
Getting started together

We start with the work that actually happens today.

Together with management and staff, we look at real cases: where is information searched for, copied across, checked or chased up? Where do queries, waiting times and errors come from?

Done by hand until yesterday
  • Matching delivery notes
  • Typing up timesheets
  • Checking invoices one by one
  • Booking goods receipts
  • Preparing payment reminders
Managed AI Automation Platform

You see what runs automatically and where your team needs to decide.

Concept view of the Managed AI Automation Platform showing automated cases, open decisions and running automations
Concept view with demonstration data. The figures shown are not a customer reference.
  1. 1What was done automatically.
  2. 2Where your team needs to decide.
  3. 3What isn’t running as expected.
  4. 4What measurable benefit the automations deliver.
AI-embedded Systems Engineering

We build systems with AI built in from the start.

Use the standard, add what is specific to you: we build on open-source platforms and extend them with real code for whatever your way of working requires. Data model, rules, permissions down to field level, migration with history.

AI tools such as ChatGPT or Claude access the system through controlled interfaces: business actions instead of free write access, checked against roles and rules, every change in the audit log.

  • Open foundationOpen-source standard systems, in the cloud or on your own EU infrastructure.
  • Controlled AI accessMCP and OAuth, role and field permissions, business actions with an audit entry.
  • Migration with historyReplace existing systems, with a test run and reconciliation before the switch.
  • Operation and developmentMonitoring, updates, integration monitoring. Further development as a rolling three-month pool instead of a monthly allowance.
Prozessfuchs Studio · Requirements Engineering

Requirements in one place, decisions you can trace.

Prozessfuchs Studio is how we manage requirements across the whole project. The requirements document and what we observe in day-to-day work are turned into a structured catalogue and confirmed together. Every requirement carries a source, owner, priority, status and acceptance criterion.

We present open questions one at a time: with a recommendation, alternatives and the requirements affected. You decide. The catalogue carries through implementation, acceptance, go-live and operation.

Less reworkUnclear points are decided before implementation, not discovered at acceptance.
A well-prepared go-liveCoverage per phase is visible at any time: accepted, tested, open.
Predictable costsA frozen baseline is the basis of the estimate. Anything after that is visible as a change.
  1. 1 · WorkshopStructure and confirm requirements

    The requirements document and observations from daily work are turned into a catalogue, completed and confirmed together.

  2. 2 · CatalogueRequirement with source, owner, acceptance criterion

    Every line can be traced back to the sentence in the source document.

  3. 3 · DecisionPresent open questions one at a time

    Recommendation, alternatives, affected requirements. You decide.

  4. 4 · BaselineFreeze the baseline and build

    Changes go through change requests; acceptance, go-live and operation work from the same catalogue.

Sovereign AI Factory · Private Frontier AI

Frontier AI under your control.

Your own AI factory for coding, copilots and internal applications: privately operable, model-agnostic and compatible with your existing AI landscape.

01

Private AI

  • Open-weight models on dedicated inference hardware
  • Your own or a rented data centre
  • Prompts, documents and source code stay within the defined operating boundary
02

Free choice of models

  • OpenAI-compatible APIs
  • Model evaluation and approval
  • Different models running in parallel
  • No lock-in to a single vendor
03

Your existing tools

  • Coding tools, chat interfaces and agents via standard interfaces
  • Private chat and copilot environment for business users
  • Existing ways of working stay intact
04

Predictable AI costs

  • Your own capacity instead of pay-per-request pricing
  • More usage raises utilisation, not the bill
  • The pilot shows the volume at which it pays off
Explore Sovereign AI
Confidential Analysis

What’s in 30,000 documents shouldn’t stay buried under 30,000 documents.

Thousands of PDFs, presentations, emails or project files. One specific question. We build a sealed-off analysis environment for it, structure the material and search for the relevant connections using private AI models and human review. Every material finding leads back to its source.

Explore Confidential Analysis
Analysis trace
  1. 31,842documents supplied
  2. 8,116possible matches
  3. 684relevant passages
  4. 37cases for review
  5. 12confirmed findings
Source trail

Finding 07→Contract_2022_04.pdf→Page 18→Section 4.2

Concept view with demonstration data. Not a customer reference.
  1. 1 / 4Understand the corpus

    File types, duplicates, scans, unreadable files: we record what is actually there.

  2. 2 / 4Analyse

    Extract, classify, relate. AI searches for the project’s questions; fixed rules take over where they are more reliable.

  3. 3 / 4Verify

    Findings are checked against their sources and can be confirmed by your team or our analysts.

  4. 4 / 4Hand over and delete

    You receive the results and a resumable project state. Then our working environment is dismantled and deleted.

About Prozessfuchs

We build and run what goes live in your business.

Together with you, we identify and automate your business processes, monitor live workflows, handle incidents and adapt them when software, rules or ways of working change.

Fabian Kübler
Lars Frobel
Maximilian Milde

Show us a task that costs you time every week.

Thirty minutes are enough for a first conversation. Afterwards, we’ll tell you whether a closer look is worthwhile.

hallo@prozessfuchs.de

One sentence about the task is enough. We reply within two business days.