Document automation in logistics: from 4 hours to 30 minutes per shipment

The problem

For every shipment, an operator had to prepare a standardised set of documents, including the CMR and a packing list with attachments. Incoming files arrived as an unsorted collection in several languages and writing systems. The operator sorted them by hand, extracted the data (manually, or with a general chat tool that often got it wrong), found the matching codes, and filled in several document sets per shipment. That took around 4 hours per shipment, every day, with no days off. Misreadings in addresses or easily confused characters forced re-checks, and one person physically could not keep up with the load.

The solution

Pinpoint AI built the automation in its simplest effective shape: a set of workflows, skills, and agents inside Claude Cowork rather than a heavy system engineered from scratch. That made it faster to put in place and cheaper for the client. The deterministic system does all the document preparation, but final approval before anything is sent stays with the operator.

How it works

1. The operator uploads every incoming file to a designated Google Drive folder.
2. The system classifies each file by its content rather than its filename, and groups documents by what they belong to.
3. Using computer vision, the AI extracts data from documents in different languages and writing systems. Every field gets a confidence level. Where confidence is low, the field is flagged for checking rather than filled in on a guess.
4. Cross-checks between documents surface inconsistencies, such as easily confused characters or identifiers that do not match, and report them to the operator.
5. The AI fills the document templates and prepares the finished sets.
6. The system writes a summary row into a shared tracking sheet and keeps an audit record linking every field back to its source.

Results

• Around 8 times faster: manual preparation per shipment fell from 4 hours to roughly 30 minutes.
• Around 66 hours freed a month, which the team can spend on other work.
• Fewer errors than manual work: automated extraction and cross-checks catch inconsistencies a tired person can miss, which cuts the number of re-checks.
• Under 1 euro per shipment in API usage: predictable, low running costs.
• The operator keeps full control: every field stays editable, and a low confidence level is clearly highlighted.