From paper form to finished dataset
A waste company received 8,000 completed paper forms from a user survey at several recycling sites. We kept the paper and built an AI workflow that cut the data entry by more than 90 per cent.
Paper worked well
The paper forms were easy to hand out and produced plenty of responses. But afterwards every tick and every handwritten comment had to be typed into Excel. At about one hour per 100 forms, the 8,000 forms came to roughly 80 hours of manual data entry.
AI read the forms
An AI workflow is a process with fixed steps in which AI handles one of them. Here, that step was the reading. An AI model read the ticks and the handwritten comments on the scanned forms, and a program gathered the answers in an Excel workbook with one row per form.
An AI model can misread. So the program flagged the forms where more than one box was ticked for the same question, or where the reading failed. We reviewed those and corrected unclear ticks and comments before the data was used.
80 hours were cut to under seven
The program could process 100 forms in about five minutes. Across all 8,000 forms, the data entry itself fell from roughly 80 hours to under seven.
Once the data was in place, we analysed the responses and wrote the report, with results for each recycling site and a comparison with the previous survey.
From form to report
The images follow the work from the completed paper forms to the Excel file and final report.
