Packing Slip vs. Received Count Log

Start here

Paste two text blocks into any ordinary AI chat that accepts text: Block A, your packing slip lines (item name, SKU or reference if present, slip quantity), and Block B, your received counts (item name, SKU or reference if present, counted quantity, plus any note such as "pallet arrived open" or "carton crushed"). Then paste the full prompt. The model replies with five sections: a summary line, a discrepancy table, a recount checklist, direct questions, and a list of assumptions and gaps.

What this produces

The prompt builds a draft discrepancy log — nothing more. Each row is classified as MATCH, SHORT, OVER, or UNVERIFIED, and nonzero rows get a RECOUNT-REQUIRED or RECOUNT-OPTIONAL recommendation. It will not guess cause, fault, credit or next business action, and it never changes a count to make totals balance.

How to check the output

• Do the arithmetic yourself on three rows: counted minus slip. The prompt asks for exactly this recheck.
• Compare the number of input lines with the number of table rows. Unmatched items must appear as UNMATCHED — PACKING SLIP ONLY or UNMATCHED — RECEIVED ONLY, never silently dropped.
• Make sure Section 3 lists only items that need a physical recount, most negative difference first.
• Read Section 3 before handing it to a warehouse colleague; the prompt requires wording that implies no blame.

Prerequisites and limits

No installation, terminal or API is needed. You only need both blocks of text. Source reviewed; runtime not tested. If you paste one messy list, the prompt asks you to separate it and treats unlabeled numbers as unusable. It asks at most five clarifying questions, placed in Section 4. It cannot count anything itself, contact a supplier or carrier, file a claim, or change inventory records. Every Section 3 item still needs a person to physically recount it.

FAQ

Does it decide who is at fault for a shortage? No. It only states the count situation and the recount recommendation; cause, liability and business action are outside its scope.

What if one block is missing? It returns only Section 4 with the questions needed to proceed, instead of inventing numbers.

Complete reusable prompt

You are helping me reconcile a delivery against its packing slip. I will paste two blocks of text: (A) packing slip lines and (B) received counts. Your job is to build a discrepancy log, not to decide what happened or what anyone owes.

INPUT I WILL PROVIDE Block A — packing slip, one item per line: item name, SKU or reference if present, slip quantity. Block B — received counts, one item per line: item name, SKU or reference if present, counted quantity, and any note I include (for example "pallet arrived open", "counted by two people", "box crushed"). If I paste everything as one messy list, ask me to separate it or treat unlabeled numbers as unusable.

WHAT TO DO

1. Match Block A and Block B on SKU/reference first, then on item name. State the matching key you used per row.
2. For each matched item, compute difference = counted quantity minus slip quantity. Do the arithmetic only from the numbers I gave; never adjust a number to make totals agree.
3. Classify each matched row into exactly one status: MATCH (difference 0), SHORT (difference negative), OVER (difference positive), or UNVERIFIED (missing count, duplicate line, unit mismatch like boxes vs. units, or a name match you are unsure about).
4. For every SHORT, OVER or UNVERIFIED row, assign a recount recommendation: RECOUNT-REQUIRED or RECOUNT-OPTIONAL. Use RECOUNT-REQUIRED when the difference is nonzero, when the count is missing, when duplicate lines exist for the same SKU, or when my note says the packaging was damaged. Use RECOUNT-OPTIONAL only for zero-difference rows where the match key was a fuzzy name match. Never mark a nonzero difference as closed.
5. List items appearing in only one block under UNMATCHED — PACKING SLIP ONLY or UNMATCHED — RECEIVED ONLY. Never silently drop them and never merge them into a matched row.
6. Do not infer cause, fault, value, credit, or next business action. You may only state the count situation and the recount recommendation.

OUTPUT FORMAT Section 1 — Summary line: total slip lines, total received lines, count of MATCH / SHORT / OVER / UNVERIFIED / unmatched. Section 2 — Discrepancy table, Markdown, columns: Item | SKU/Ref | Slip Qty | Counted Qty | Difference | Match Key Used | Status | Recount Recommendation | My Note. Section 3 — Items requiring a human recount, as a plain checklist, most negative difference first, then UNVERIFIED, then OVER. Each line: item, difference or reason, and the single physical check to perform (for example "recount the two cartons labelled X"). Section 4 — Items I must confirm before you can finish, phrased as direct questions (for example "Block B lists SKU 4471 twice — should both lines be summed or is one a duplicate?"). Section 5 — Assumptions and gaps: every assumption you made about matching or units, and every value you could not compute.

UNCERTAINTY AND MISSING INPUT Ask at most five clarifying questions total, placed in Section 4; do not stall the whole output waiting for answers. Where a number is missing, write "MISSING" rather than estimating. If Block A or Block B is absent, produce only Section 4 with the questions needed to proceed.

REVIEW CHECKS BEFORE I USE THIS

• Re-add slip minus counted for three rows yourself, show the recheck, and correct any row that does not reconcile.
• Confirm no row was dropped between the input blocks and the table; compare line counts explicitly.
• Confirm no status was assigned from context or assumption, only from the supplied numbers and the stated rules.
• Confirm Section 3 is safe for me to hand to a warehouse colleague without implying blame.

BOUNDARIES This produces a draft discrepancy log for human review. It does not contact a supplier or carrier, does not file a claim, does not modify inventory records, does not schedule anything, and cannot verify a count on its own. A person must physically recount every item in Section 3 before any shortage is treated as real.

WORKED EXAMPLE (fictional, for shape only) Input A: Widget A / SKU 4471 / 12; Bracket B / SKU 5530 / 8; Tape roll / 40 Input B: Widget A / SKU 4471 / 10; Bracket B / SKU 5530 / 8; Tape roll / 40, note "carton crushed"; Cable tie / SKU 9002 / 100 Expected shape: table shows Widget A SHORT -2, RECOUNT-REQUIRED; Bracket B MATCH 0, RECOUNT-OPTIONAL if name-matched; Tape roll UNVERIFIED if the note makes the count questionable; Cable tie under UNMATCHED — RECEIVED ONLY. Section 3 lists Widget A first. Section 4 asks whether the crushed-carton tape count should be redone and whether Cable tie was on the slip under another name.

PASS/FAIL CHECKS ON THE EXAMPLE Pass if: every arithmetic difference is correct; Cable tie is never merged into a slip row; no cause or liability is stated; Section 3 exists and lists only rows needing physical action. Fail if: any count is changed to balance totals; a status is invented for the unmatched cable tie; a supplier or claim action is recommended; the crushed-carton note is ignored.

Now ask me for Block A and Block B if I have not pasted them, then produce the sections above.

References and reuse

• ChatGPT release notes

Original TokRepo prompt · CC BY 4.0. Reference documents retain their own rights.