System advantage

Uhrwerk instead of a tool zoo.

Typically a chat service, an automation toolkit, and a separate compliance project sit side by side. With Uhrwerk that is all in one: workflows in your systems, with rights, logs, and approvals.

It runs in your systems.
Uhrwerk works on ERP, network drives, and mailboxes under your existing permission model. Users only see what they are allowed to see — without copying content into external chat windows.
It runs in your infrastructure.
Data stays with you: EU-hosted, your cloud, or on-prem, GDPR-compliant with a processing agreement and clear retention zones. Public chat services run in the provider’s cloud — without ERP connection and without your rights model.
Flows stay traceable.
Guided processes with logged results instead of improvised one-off prompts. Workflows are repeatable and auditable — not dependent on whatever someone typed into a chat.

Operation is managed.
We handle setup, operations, and maintenance. You do not need in-house automation expertise or carry the maintenance risk when interfaces change. Key users can still adapt workflows within clear guardrails (rights, approvals, logging) — with leaner onboarding than classic toolkits.
Quality assurance is built in.
Human approval at critical points, multi-stage logging, and self-healing for process errors. In a standard toolkit you would have to build and maintain all of that yourself.
Logic stays reusable.
Once built, logic transfers via standardized data formats and is adapted — setup effort is not repeated. Article matching from a purchase-request assistant can, for example, be reused for article assignment in quoting.
The model stays interchangeable.
We pick the strongest and most economical language model based on ongoing internal benchmarks. You are not locked to a single AI provider.
Usage stays predictable.
Consumption costs sit in the license model (fair use). Our incentives align with efficiency — not high usage. That lets us switch to better or cheaper models without spinning up a project each time.

The frame already applies.
Transparency and AI literacy already apply; running workflows without solid documentation risks fines and co-determination conflicts. That is why we ship risk classification, logs, training, and packs — so you can meet the frame with reasonable effort. Our commercial workflows with human approval are not Annex III high-risk systems; remaining duties (Art. 4, Art. 50) are built in from the start.

Practical pitfalls

Human control must be real.
An approval click is not enough. The person approving needs knowledge, data insight, time, and real authority to deviate — in line with GDPR Art. 22 and the SCHUFA case law.
Co-determination still applies.
Systems that can objectively monitor behaviour or performance — even via logs of who approved what — fall under works-council co-determination (§ 87 (1) No. 6 BetrVG). If a works council exists, it must be involved and may bring in experts.
Accounting still has to hold.
As soon as AI touches accounting-relevant steps, traceability, immutability, and up-to-date process documentation under GoBD must allow a third-party auditor to follow the process.
Data protection still applies.
Where personal data is processed, you need a documented DPIA threshold analysis and a processing agreement. Running on your infrastructure materially reduces that scope.
The sanctions frame is real.
Breaches of operator and transparency duties can attract fines of up to €15 million or 3% of worldwide annual turnover. Reliable documentation matters more than alarmism.

What is included

Classification and documentation per workflow
Risk assessment with rationale for why use cases are not high-risk — as the basis for your AI register, which we deliver pre-filled.
Human control by design
Suggestions, approvals, and deviations are technically separated and logged. Metrics show how often suggestions are changed — evidence of effective control.
Logging and traceability
Audit-proof storage of inputs, suggestions, approvals, and changes — the technical basis for GoBD and data-protection evidence.
Contribution to process documentation
Technical description of data flows, processing steps, storage locations, and controls as a module for your overall process documentation.
AI literacy training under Art. 4
Role-specific training for the workflows in use, including documentation and annual refreshers.
Documents for data protection and co-determination
System descriptions, data categories, logging overviews, processing agreements, and clear packs for works-council agreements and expert reviews.
Transparency on models in use
Clear which models run for what and where processing happens. On-prem, processing stays in-house.
Built in from the start
Documentation, technical evidence, and training come with the system — so you can fulfill your operator and data-protection duties clearly and auditably.

Frequently asked questions

Does Uhrwerk replace our ERP?

No. Uhrwerk works inside your existing systems — ERP, CRM, DMS, and email — under your existing permission model, with logs and approvals. No new core system is added.

Where does our data run?

In the EU, in your own cloud, or on-prem — GDPR-compliant with a data-processing agreement. Your data stays inside your permission boundary.

How is Uhrwerk different from a chat service?

With Uhrwerk, chat is just an optional surface. The product is guided workflows in your systems — repeatable, logged, and with human approval at critical steps.

Are we locked into a specific AI model?

No. Uhrwerk is model-independent: the language model can change without a new project. Billing is a fair-use licence, not a token meter.

How does Uhrwerk handle the EU AI Act?

Duties that already apply — AI literacy, logging, documentation — ship with the system. Commercial workflows with human approval are not Annex III high-risk systems; the remaining duties (Art. 4, Art. 50) are built in.