Converting handwritten time sheets into ERP data
Koninklijke Vezet now reads handwritten time sheets from 80 production lines automatically, saving its time registration team an estimated 10 hours a day.
Koninklijke Vezet processes vegetables and fruit into fresh-cut products, soups and meal salads for Albert Heijn and Delhaize. Production runs on 80 lines, with up to two shifts a day. We worked with the time registration department and the IT team.
Challenge
Every day, line operators fill in a paper time sheet by hand. They record when machine setup started, when each product run began, how long each activity took and how many people worked on the line. With 80 lines and up to two shifts a day, the paper piles up fast. The time registration department retyped every sheet into Vezet's ERP system, which took about 10 hours of work every day. Reading handwriting led to typing errors, and moving data from paper to system involved many unnecessary manual steps. The result was a slow process with a constant risk of mistakes in registered hours.
Solution
We built a system that scans paper time sheets in bulk and reads the handwriting with two separate AI models. When both models read the same value, it goes straight into the ERP. When they disagree, an employee checks that field in a dashboard.
- Bulk scanning: uploads stacks of paper time sheets in one go.
- Dual reading: two text-recognition models each read every field independently.
- Automatic registration: sends values both models agree on to Vezet's custom-built ERP system.
- Review dashboard: shows only uncertain fields, so staff can confirm or correct them quickly.
No uncertain value enters the ERP without a human check.
Research first, then build
We tested handwriting recognition before writing production code.
- Process analysis We mapped how time sheets moved from the production line into the ERP, and what the department needed.
- Model research We tested which models read handwritten text best and measured their performance. This research was completed before implementation started.
- Build We built the scanning pipeline and the connection to Vezet's ERP system.
- Dashboard design We made the review and correction steps as easy as possible for the time registration team.
Reimer KoopalAI Engineer
Titus DriessenAssociate AI EngineerImpact
The time registration department no longer retypes time sheets by hand. The system reads and registers most fields automatically, and only a small share of sheets still needs a manual check. That check happens in a simple dashboard, not through re-entry. The department saves an estimated 10 hours of data entry per day. Fewer manual steps also leave less room for typing errors in registered hours. Vezet's IT team especially valued the scientific research behind the choice of models. The system is live and in use by the time registration department.
More cases
At AdamI, we transform businesses by using the technology of artificial intelligence
