Choose one of three approaches. The steps for that option appear below.
Choose an option above. On a phone, swipe the tabs left or right to see all options →
ChatGPT pairs claim lines with receipts, drafts the completeness report. You correct the pairing and approve; the approver confirms thresholds.
Get ready
Pair the receipts
Correct the pairing
Draft the report
Confirm thresholds
Approve
Use: You
The claim export with 45 lines, the 38 receipt images and the policy no-receipt rules and thresholds.
Check the export is complete and you may upload it; keep the originals.
Use: ChatGPT
Upload the claim export and receipt copies; ask for a table pairing each claim line with its receipt, recording merchant, date and total.
What to paste into the tool
Task: Pairing table with content verdicts. Pair each of the claim lines with a receipt image if one exists and record merchant, date and total from the image. Keep three verdicts separate: receipt present and readable, receipt present but unreadable, and no receipt. Do not treat unreadable as missing. Write UNREADABLE with the reason where a value is not legible on the image.
Use: You
Open every receipt marked missing and look for it in your files, then correct the pairing table.
A receipt marked missing must be genuinely absent, not misfiled.
Use: ChatGPT
In the same chat, paste the checked table and the policy rules and ask for the completeness report with the policy section cited per threshold.
What to paste into the tool
Task: Draft completeness report: lines without receipts, receipts missing merchant, date or total, and lines over the no-receipt threshold. List the lines without receipts, the receipts missing merchant, date or total, and the lines over the no-receipt threshold, citing the policy section for every threshold used. Keep three verdicts separate: receipt present and readable, receipt present but unreadable, and no receipt. Do not treat unreadable as missing.
Use: The approver
Check each threshold applied against the current policy and confirm the policy section is cited for every rule used.
Use: You
Read the missing-items list, find any receipts that were misfiled, note which lines still need a receipt and approve the final report.
Chasing missing receipts is the claimant's action outside this draft.
You pair lines with receipt files; Claude reads the receipts in batches. The approver applies policy thresholds in a spreadsheet.
Pair the files
Read in batches
Sample-check
Apply thresholds
Finish
Use: You
The claim export with 45 lines and 38 receipt images. Pair each claim line with a receipt file and list lines without one.
Receipts marked missing must be genuinely absent, not misfiled.
Use: Claude
Upload the receipts in batches of up to 19 per chat with the pairing list; ask for merchant, date and total for each receipt.
What to paste into the tool
Task: Pairing table with content verdicts. For each receipt in this batch record merchant, date and total next to its line from the pairing list. Use three verdicts: readable, unreadable with the reason, and no receipt for the lines with no file. Do not apply the policy.
Use: You
Compare a sample of amounts and dates with the receipt images, and check every receipt marked unreadable.
Investigate any mismatch against the image before going on. The vendor caps each chat at 20 files.
Use: The approver
Apply the no-receipt rule and threshold values from the current policy to the pairing table in a spreadsheet and list exceptions with the policy section.
Use: You
Read the missing-items list, find any receipts that were misfiled and note which lines still need a receipt.
Chasing missing receipts is the claimant's action outside this draft. A receipt marked missing must be genuinely absent, not misfiled.
Gemini turns the policy into a checklist, checks line by line. You and the approver confirm each stage before next.
Get ready
Draft the checklist
Confirm the checklist
Pair and record
Correct the pairing
Draft the report
Confirm thresholds
Approve
Use: You
The claim export with 45 lines, the 38 receipt images and the policy no-receipt rules and thresholds.
Check the export is complete and you may upload it; keep the originals.
Use: Gemini
Upload the policy and ask for a checklist of required receipt content and the no-receipt thresholds, with the policy section quoted for each rule.
What to paste into the tool
Task: Policy checklist with quoted sections. From the policy document, list each rule for required receipt content and each no-receipt threshold, with the section number and quoted wording for each. Do not interpret unclear wording; write UNCLEAR WORDING where the policy is ambiguous.
Use: The approver
Read the drafted checklist against the current policy, correct it and confirm it before any checking step uses it.
Use: Gemini
Upload the claim export and up to nine receipt copies at a time; ask for pairing and content verdicts using the checklist.
What to paste into the tool
Task: Pairing table with content verdicts. Pair each claim line in the batch with a receipt image and record merchant, date and total. Keep three verdicts separate: receipt present and readable, receipt present but unreadable, and no receipt. Do not treat unreadable as missing. Check each receipt against the checklist content rules only.
Use: You
Open every receipt marked missing and look for it in your files, then correct the pairing table.
Unclear wording goes to the approver for a ruling.
Use: Gemini
Paste the checked table and the checklist and ask for the completeness report with the section cited for each threshold.
What to paste into the tool
Task: Draft completeness report: lines without receipts, receipts missing merchant, date or total, and lines over the no-receipt threshold. Apply only the checklist thresholds. List lines without receipts, receipts missing merchant, date or total, and lines over the no-receipt threshold, citing the checklist section for each. Keep three verdicts separate: receipt present and readable, receipt present but unreadable, and no receipt. Do not treat unreadable as missing.
Use: The approver
Check each threshold applied against the current policy and confirm the policy section is cited for every rule used.
Use: You
Read the missing-items list, find any receipts that were misfiled, note which lines still need a receipt and approve the final report.
Chasing missing receipts is the claimant's action outside this draft. A receipt marked missing must be genuinely absent, not misfiled.
Best when: Many images; one chat holds pairing and report
What it does: ChatGPT pairs and reports; you check, approve
Strong at: The checked pairing table feeds the report directly
Limit: Image values may extract inaccurately; check verdicts
Free plan: Yes, limited (uploads and analysis capped)
Best when: You pair by file; the approver applies thresholds
What it does: Claude reads receipts; you and the approver decide
Strong at: No AI step applies a policy threshold
Limit: Pairing by file name adds manual work
Free plan: Yes
Best when: Rules fixed in a checklist before any AI step
What it does: Gemini checks line by line; you and approver confirm
Strong at: Both checks read from the same stated rules
Limit: 38 images span four batches, combined by hand
Free plan: Yes
Before you paste into any AI tool, remove any confidential information. Put it back at the end.
Deciding between ChatGPT and Claude for this? See the comparison.
Never paste credentials or full payment details.
Confirm that receipts marked missing are genuinely absent, not misfiled.
Confirm threshold rules from the current policy; chasing missing receipts is the claimant's action outside this draft.
Every claim line has a receipt verdict; every receipt has a content verdict; thresholds applied come from the stated policy; the claimant reviews the missing-items list.
Check the result against your original inputs. Approve it before sharing or acting.