Inventory Exception Agent Workflow

Goal

Generate a unified exception report that flags low-stock and oversell risks, recommends reorder quantities per supplier, drafts supplier emails, and provides a verification checklist before purchase orders are issued.

Inputs to gather

• Shopify product inventory export file(s) (provides on-hand counts, oversell policy).
• Shopify order export file(s) (reveals sales velocity and open commitments).
• Warehouse or cycle-count file(s) (actual on-hand and damaged/held quantities).
• Refund/return log(s) (adjusts velocity and available stock).
• Supplier spreadsheet(s) with lead time, MOQ, case pack, price, currency (needed for reorder math and email drafts).
• Optional: open PO/inbound files (offset reorder qty), SKU crosswalks, bundle BOMs (translate kits to components).
• Parameters: safety stock days, review period days, sales-velocity lookback windows, receiving buffer days (drive threshold calculations).
• Business rules: pooled vs. per-location inventory, excluded SKUs, supplier holidays/closures (refine recommendations).

Before doing any work, ask the user for these inputs in ONE message. Skip anything they already provided. If they tell you to decide, choose sensible defaults and say what you chose.

Workflow

1. Confirm scope and collect any missing inputs or parameter preferences from the user; suggest defaults if they defer.
2. Ingest all supplied files using the Read tool. Validate required columns; if any are missing or ambiguous, pause and request clarification.
3. Clean and normalize data:
   a. Standardize SKU formatting and apply any provided crosswalks.
   b. Expand bundle demand into component SKUs via BOMs.
   c. Convert currencies to a base currency where costs are needed.
4. Build a unified daily SKU-level demand table from order exports, subtracting cancelled items and adjusting for refund disposition.
5. Merge warehouse counts, Shopify availability, inbound POs, and supplier attributes into a master inventory table (per SKU, per location if required).
6. Calculate sales velocities (7-day, 30-day, optional 90-day) and choose the governing velocity per SKU using the rules in the source workflow. Adjust for stockouts where detectable.
7. Compute available-to-promise (ATP), reorder point, target stock, suggested reorder quantity, and projected depletion date, all respecting MOQs, case packs, lead times, and safety stock parameters.
8. Flag exceptions:
   • Low-stock alerts where ATP ≤ ROP or days-of-cover falls below lead-time + safety-days.
   • Oversell risks where ATP < 0, depletion precedes inbound + buffer, or oversell policy is enabled on high-velocity items.
   • Data-quality issues (missing lead time, unknown supplier, etc.).
9. Generate supplier-grouped reorder suggestions, including costs and projected post-reorder coverage.
10. Draft one email per supplier requesting the suggested quantities, confirming price/lead-time, and noting any special instructions.
11. Compile a run-specific verification checklist covering counts, inbound, bundles, refunds, catalog status, demand spikes, supplier constraints, policy flags, capacity, budget, and channel sync.
12. Use the Edit tool to write these files:
    • alerts_low_stock.csv
    • risks_oversell.csv
    • reorder_suggestions.csv
    • supplier_email_drafts.md
    • verification_checklist.md
    • summary.md (concise human-readable overview)
13. Present the summary to the user, highlighting the top urgent SKUs, counts of exceptions, and total suggested spend per supplier. Offer to rerun with alternate parameters if desired.

Output

Six files saved in the workspace:

1. alerts_low_stock.csv – SKU-level low-stock list with ATP, days cover, depletion date.
2. risks_oversell.csv – SKU-level oversell risks with reason codes and mitigation notes.
3. reorder_suggestions.csv – Supplier-grouped reorder quantities, MOQs, case packs, costs.
4. supplier_email_drafts.md – One well-formatted draft email section per supplier.
5. verification_checklist.md – The tailored checklist for human review before POs.
6. summary.md – Executive summary (max ~300-400 words) of the run’s key findings and next steps.