The challenge
Manually matching physical delivery notes against the original orders in goods receipt was time-consuming, resource-intensive and error-prone. Staff had no efficient way to process complex delivery documents quickly. Capturing and matching quantity deviations, quality differences and countries of origin by hand led to delays in posting, faulty documentation in the ERP system and a lack of transparency in the supply chain.
Architecture and technical implementation
To establish an automated and intelligent goods receipt, we implemented a tailored solution built on modern artificial intelligence.
- A combined visual and language model made of a vision and large language model automatically reads complex, unstructured delivery notes in the most varied layouts and interprets the relevant business data.
- An intelligent logic module compares the extracted delivery note data with the corresponding order in real time. Any deviations in quantities, quality features or countries of origin are detected immediately and presented clearly to staff for review, with a human in the loop.
- The verified and structured records flow into the existing ERP system immediately via direct API interfaces, with no manual transfer work.
- The solution continuously captures feedback from operations, learns from manual corrections and adapts individually to the specific supplier structures and needs of the customer.
Business value
- With the manual retyping gone, data quality in the ERP system rises noticeably. Downstream departments such as production and accounting can rely on valid real-time data.
- The time-consuming manual matching is almost entirely eliminated. Quantity and quality differences as well as regulatory metadata such as countries of origin for customs are documented precisely and in fractions of a second.
