In German subcontracting projects, two documents come up most often: the Freistellungsbescheinigung and the A1 certificate. The first confirms exemption from withholding tax for construction services. The second shows that an employee temporarily posted to another country remains under the social security system of the sending state, in line with EU rules described by the European Commission and EURES.
In practice, the problem is not simply having the file, but the fact that documents arrive in different versions, languages and formats, and their validity dates, data consistency and assignment to a specific subcontractor must all be tracked. This stage is well suited to AI, provided the system first extracts data from the document and only then compares it with the contractor register.
According to Minerva's documentation on offers and tender documentation, AI can read full documents, including scanned PDFs, extract requirements and compare them with the set of prepared materials, flagging gaps and points requiring clarification. The same mechanism can be applied to subcontractor documents: the system detects the document type, reads the company name, dates, validity period and identifiers, and then compares them with the supplier or subcontractor record.
According to the description of document use cases on die-ki-agentur.de, AI can also classify documents, check their completeness and detect deviations from required rules. For the Freistellungsbescheinigung, this means a simple control scheme: whether the document concerns the correct entity, whether it has not expired, whether it covers the correct service period, and whether it is attached to the correct project.
The greatest risk is trusting the reading without verifying the source. According to the law firm Luther, discussing the position of the BRAK, results generated by AI must be verified by a human to avoid errors and liability. In subcontractor documentation this matters especially, because a mistake in a number, a date or a validity status can cause costly chaos in accounting or compliance.
The practical working model is therefore as follows: AI reads the document, extracts the fields, flags deviations and creates an alert, but the decision is approved by an employee. The same applies to A1, where the system should signal a missing certificate before work begins in Germany, or a mismatch between the person on the list and the data in the document.
Lub System helps B2B companies implement AI, automation and IT solutions end-to-end - from strategy to deployment. See our services or get in touch to discuss your case.
Source: https://www.useminerva.com/pl/blog/zgodnosc-oferty-swz-ai