Claude Cowork workflow
Bill of Lading Data Extraction with Cowork
Extract and reconcile BOL fields against your TMS records, flagging missing data and discrepancies.
What this workflow does
Use Cowork to extract structured data from approved Bills of Lading: shipper, receiver, origin, destination, items, weight, charges, pickup/delivery dates, and signature status. Cowork flags fields it cannot read with confidence (UNREADABLE), compares extracted data against your TMS records, and identifies mismatches. Operations staff verifies discrepancies in the source documents and updates the system of record.
This workflow is designed to save time on manual BOL data entry and catch TMS/BOL discrepancies early. Instead of staff manually typing fields from paper or PDF BOLs, Cowork extracts the data and flags confidence issues so your team can focus on reconciliation and corrections.
When to use this workflow
Use this workflow when you process BOLs daily and need to verify them against your TMS. It works best with legible, standard BOL formats. For handwritten or damaged documents, consider image cleanup or manual review first. Scale from 10 BOLs to hundreds once your team validates the workflow.
Runbook
- Gather an approved set of Bills of Lading from your document folder.
- Provide reference data: your TMS export or a manual inventory of expected shipments.
- Ask Cowork to extract: shipper, receiver, items, weight, charges, dates, signatures, and exceptions.
- Flag low-confidence or missing values; reconcile against your records; verify discrepancies in the source documents.
Real-world example
A freight company receives 50 BOLs daily from various carriers. Operations uploads 10 sample BOLs in PDF format and their TMS export. Cowork extracts shipper, receiver, weight, charges, and pickup/delivery dates from each BOL, then compares against the TMS. The workflow finds 3 discrepancies: one BOL shows 500 lbs but TMS shows 505 lbs (possible rounding or weight variance); another has a delivery date 1 day earlier than promised; a third is missing a signature scan. Operations investigates the source BOLs, confirms the weight variance is acceptable, schedules a follow-up call on the late delivery, and flags the signature for manual review. The workflow scales to 50 BOLs per day after this validation.
Copy-and-run prompt
Read all Bills of Lading in my documents folder. For each BOL, extract: shipper name and ID, receiver name and ID, origin city/zip, destination city/zip, item count, total weight, declared value, shipping charges, pickup date, delivery date, carrier, signature presence, and any exception notes. If a field is unreadable or missing, mark it UNREADABLE. Compare against my TMS reference data and flag any mismatches. Create a reconciliation table with BOL ID, extracted values, TMS values, and discrepancy status. Do not modify any records.
Human review checklist
- Review every UNREADABLE field in the source BOL document.
- Verify all dates, shipment IDs, and party identifiers match the original paper or image.
- Investigate and resolve TMS/BOL discrepancies by checking the system of record.
- Confirm signature presence and carrier approval.
- Use reconciliation output to update any stale or incomplete TMS records.
Handling confidence and unreadable data
Cowork will extract fields it can read clearly and mark uncertain or illegible content as UNREADABLE. This prevents false precision: staff can then manually review the original BOL and fill in the value with confidence. Common UNREADABLE scenarios: handwritten signatures, faded dates, smudged addresses, or non-standard BOL formats. Always have operations staff verify critical fields (charges, delivery confirmation, exception notes) against the original BOL before relying on extracted data.
Reconciliation and TMS accuracy
BOL data extraction is only valuable if it leads to accurate TMS records. When you first implement this workflow, plan for a careful reconciliation phase. Extract 10-20 BOLs, compare the extracted data against your TMS, and investigate every discrepancy. You'll often find issues like: TMS was updated with a customer estimate instead of the actual weight; BOL shows a different delivery address than what was promised; or carrier signatures are missing from the TMS entirely. These discrepancies are not failures of the workflow; they're opportunities to fix underlying data quality issues in your TMS.
Once you've investigated the first batch and updated TMS records where needed, the workflow becomes more powerful. You now have a daily mechanism to catch TMS/BOL mismatches as they happen, before they cause downstream problems like billing disputes or missed deliveries. Set up a process where your operations team reviews the reconciliation output daily, flags anything suspicious, and either verifies the TMS record or corrects it based on the BOL.
Managing confidence and UNREADABLE fields
Cowork's UNREADABLE flag is a feature, not a limitation. It means Cowork is telling you "I don't trust this extraction." This prevents silent failures where the system confidently extracts the wrong weight or charge amount. When you see an UNREADABLE field, your staff manually reviews the original BOL and makes a decision: the field is genuinely illegible (route it to a supervisor), or it's readable but Cowork just didn't extract it correctly (fix the source data or retrain the workflow).
Track which fields are most commonly flagged as UNREADABLE. If signatures are always UNREADABLE, maybe your BOL format doesn't include them or they're in a position Cowork doesn't expect. If weights are often UNREADABLE, possibly your carriers are writing them in non-standard units. Use these patterns to either improve your source materials or adjust your workflow expectations. Over time, as your BOLs become more consistent, the UNREADABLE rate usually drops and extraction becomes faster and more reliable.
Common pitfalls
Don't: Assume Cowork extraction accuracy is 100%; always verify against the original. Don't: Upload handwritten BOLs without first testing on legible samples. Don't: Change your TMS based on extracted data without operations approval. Do: Update TMS when extraction reveals stale or incorrect records. Do: Archive the original BOLs for 90 days in case discrepancies surface later. Do: Compare extracted weight/charges side-by-side with TMS to catch mismatches early. Do: Track UNREADABLE rates and investigate trends; they tell you where your data quality needs improvement.
Frequently asked questions
- Can Cowork read BOLs accurately?
- It can extract structured fields with high confidence for clear, legible documents. Handwritten or damaged BOLs require manual inspection or image cleanup first.
- What does UNREADABLE mean?
- When Cowork cannot confidently extract a field (e.g., signature, smudged date), it flags it UNREADABLE instead of guessing. Staff reviews the original BOL and enters the value manually.
- How do we handle BOL vs. TMS mismatches?
- Cowork flags discrepancies but does not change records. Operations staff verifies the source documents, identifies the truth, and updates the system of record manually.
- Can Cowork validate charges or weights?
- It can extract and compare them side-by-side. Final authority on correctness stays with your billing and operations teams.