Receipt Finder & Organizer

Business name: Northstar Creative Studio Primary mailbox: Company Gmail inbox Finance email: finance@northstar-example.com Storage: Google Drive Root folder: Finance Records Processing mode: Hybrid Automatic-processing rule: Automatically forward and store only high-confidence business receipts from approved vendors or messages clearly associated with the company payment card. Review rule: Queue personal, ambiguous, unreadable, password-protected, duplicated, refunded, or incomplete documents. Search period: Process new messages daily. Perform an initial historical backfill from January 1, 2026. Document types: Receipts Invoices Payment confirmations Subscription charges Refunds Credit notes Travel receipts Advertising receipts Software invoices Professional-service invoices Approved vendors: Adobe Google Workspace Microsoft Amazon Business Canva Figma OpenAI Dropbox Slack Notion Zoom Meta Ads Google Ads Local internet provider Office-supply vendors Approved travel vendors Known recurring subscriptions: Adobe — monthly Google Workspace — monthly Microsoft 365 — monthly Canva — monthly Figma — monthly Dropbox — monthly Slack — monthly Notion — monthly Zoom — monthly Business-identification signals: Company billing name Company tax identifier Company billing address Company card ending in approved masked digits Business email address Approved vendor Project or client reference Expense-category keyword Monthly folder structure: Finance Records → Year → MM - Month → Receipts → Invoices → Refunds and Credits → Exceptions → Processing Log Folder-date policy: Use transaction date when clearly available. Otherwise use invoice issue date. If neither is reliable, use the email-received month and mark Requires Date Verification. File-name format: YYYY-MM-DD - Vendor - Amount Currency - Document Type - Reference.ext Forwarding subject: Receipt | YYYY-MM | Vendor | Amount Currency | Reference Forwarding body: Vendor Transaction date Amount and currency Document type Expense category Storage destination Business-purpose confidence Exception note where applicable Expense categories: Software Advertising Office Supplies Travel Internet and Communications Professional Services Education Equipment Other Business Expense Duplicate rules: Use source message ID where available. Compare attachment hash where supported. Compare vendor, amount, currency, transaction date, and invoice or order number. Do not forward a duplicate automatically. Refund policy: Store refunds and credit notes in Refunds and Credits. Attempt to link them to the original expense. Do not count them as new expenses. Late-arrival policy: Store the document in the transaction month. Mark it as a late arrival in the monthly processing log. Missing-receipt monitoring: At monthly close, check whether expected recurring-vendor receipts were found. Create a missing-receipt report. Sensitive-data rules: Do not forward unrelated email-thread content. Do not expose complete card numbers or account credentials. Preserve original documents. Mask sensitive payment references in reports where appropriate. Need output: Complete automation architecture Search strategy Receipt qualification rules Business-purpose classification Monthly folder architecture File-naming specification Forwarding specification Duplicate controls Refund handling Late-arrival handling Subscription tracking Exception workflow Audit log Historical backfill plan Monthly close plan Reconciliation framework Acceptance criteria Regression tests QA checklist Risk assessment Special constraints: Do not automatically process personal purchases. Do not delete source emails. Do not delete original attachments. Do not repeat a finance forward after partial failure. Do not convert currency. Do not invent tax deductibility. Do not claim reconciliation is complete when statement data is missing.

=== RECEIPT FINDER & ORGANIZER ARCHITECTURE === Business: Northstar Creative Studio Processing Mode: Hybrid Primary Mailbox: Authorized company Gmail inbox Finance Destination: finance@northstar-example.com Storage Root: Finance Records === EXECUTIVE SUMMARY === The workflow will search new email daily and perform a controlled historical backfill beginning January 1, 2026. High-confidence business receipts will be: 1. Qualified as a supported expense document. 2. Classified as business-related. 3. Checked for duplicates. 4. Assigned to the correct transaction month. 5. Renamed using the approved naming pattern. 6. Stored in the correct monthly folder. 7. Forwarded once to finance. 8. Written to the processing log. Ambiguous, personal, unreadable, duplicated, password-protected, incomplete, refunded, or conflicting documents will enter the review queue. Source messages and original attachments will not be deleted. === SEARCH STRATEGY === Search Signals: Document Keywords: - receipt - invoice - tax invoice - payment confirmation - paid - order confirmation - subscription - renewal - refund - credit note - billing statement Attachment Signals: - PDF - image receipt - document attachment - structured HTML receipt Sender Signals: - approved vendor domains - billing aliases - payment processors - subscription services Content Signals: - amount - currency - invoice number - order number - transaction date - payment method reference - company billing name - company billing address Exclusions: - newsletters - promotional offers - abandoned-cart messages - shipping-only updates - product announcements - password-reset messages - unrelated account alerts - personal purchases - messages already marked as successfully processed === RECEIPT QUALIFICATION === High Confidence: A document is high confidence when: - the sender is an approved vendor or verified billing domain - a receipt or invoice document is present - an amount and currency are visible - a transaction or invoice date is visible - the company identity or approved payment method is present - the message represents a completed charge, invoice, refund, or credit - no strong personal-use signal is present Medium Confidence: - probable receipt - amount is visible - vendor is identifiable - business purpose is plausible - one important field is missing or unclear Action: Queue for review. Low Confidence: - marketing message - incomplete order - no completed-payment evidence - personal purchase signal - unrelated attachment - unsupported file - unreadable document Action: Skip or queue according to risk. === BUSINESS-PURPOSE CLASSIFICATION === Business Confirmed: Approved vendor, company billing identity, approved company card, business email, project reference, or approved expense category. Personal: Personal billing identity, personal shipping purpose, household purchase, or user-defined personal vendor. Ambiguous: Insufficient evidence to distinguish business from personal. Rule: Ambiguous documents are never forwarded automatically. === DOCUMENT TYPES === Receipt: Evidence of completed payment. Invoice: Request for payment or formal billing document. Payment Confirmation: Evidence that a payment was processed. Subscription Charge: Recurring service charge. Refund: Money returned for an earlier expense. Credit Note: Formal reduction or reversal. Cancellation: Document indicating that a transaction or service was cancelled. Exception: Document requiring manual interpretation. === DATE POLICY === Primary: Transaction date Secondary: Invoice issue date Fallback: Email received date Fallback Status: Requires Date Verification Late Arrival: A document received after its transaction month is filed in the transaction month and marked Late Arrival. === MONTHLY FOLDER STRUCTURE === Finance Records └── 2026 └── 07 - July ├── Receipts ├── Invoices ├── Refunds and Credits ├── Exceptions └── Processing Log Folder Rules: - Create the year folder when absent. - Create the month folder when absent. - Do not create repeated month folders with alternate names. - Store refunds and credit notes separately. - Store unresolved documents in Exceptions. - Preserve a processing log for every month. === FILE-NAMING STANDARD === Format: YYYY-MM-DD - Vendor - Amount Currency - Document Type - Reference.ext Example: 2026-07-14 - Adobe - 59.99 USD - Invoice - INV-84721.pdf Fallback Without Reference: 2026-07-14 - Adobe - 59.99 USD - Invoice.pdf Fallback Without Verified Amount: 2026-07-14 - Adobe - Amount Unverified - Invoice - INV-84721.pdf Rules: - Sanitize unsafe characters. - Preserve the original extension. - Do not overwrite an existing unrelated file. - Use a duplicate-safe suffix only after duplicate analysis. - Keep names understandable outside the email system. === FORWARDING SPECIFICATION === Recipient: finance@northstar-example.com Subject: Receipt | 2026-07 | Adobe | 59.99 USD | INV-84721 Body: Vendor: Adobe Transaction Date: July 14, 2026 Amount: 59.99 USD Document Type: Invoice Expense Category: Software Storage Destination: Finance Records/2026/07 - July/Invoices Business-Purpose Confidence: High Source: Original company mailbox message Attachments: Original verified invoice Rules: - Forward once. - Include only relevant context. - Do not forward unrelated thread content. - Preserve the original document. - Record the resulting forward status. - Do not repeat the forward when a later storage or logging action fails. === DUPLICATE CONTROL === Primary Identifiers: - source message ID - attachment hash where supported - invoice number - order number - transaction number Secondary Comparison: - vendor - amount - currency - transaction date - attachment name Duplicate Statuses: - Exact Duplicate - Probable Duplicate - Related Document - Resent Document - Not Duplicate Exact Duplicate: Skip forwarding and storage. Record the existing document reference. Probable Duplicate: Queue for review. Related Refund or Credit: Link to the original expense rather than treating it as a duplicate expense. === IDEMPOTENCY RECORD === For every document, record: - source message identifier - attachment identifier - attachment hash where supported - vendor - amount - currency - transaction date - reference - destination file - forwarding status - storage status - log status - final processing status A completed record must not be processed again. === REFUND AND CREDIT HANDLING === Refund Flow: Refund Discovered → Identify Original Vendor → Match Amount or Reference → Link to Original Expense → Store in Refunds and Credits → Forward to Finance → Mark Original Relationship → Record in Log Do Not: Count a refund as a new expense. Unmatched Refund: Store in Exceptions or Refunds and Credits according to policy and mark Original Expense Not Confirmed. === SUBSCRIPTION RECEIPT TRACKER === Vendor: Adobe Frequency: Monthly Expected Date Window: Configured by finance Expected Currency: USD Expected Amount: Verified value or approved range Current Month: Found / Missing / Ambiguous / Refunded Last Receipt: Verified document reference At monthly close: - compare expected subscriptions with discovered documents - identify missing receipts - identify unexpected duplicate charges - identify changed amounts for review - do not assume an amount change is incorrect === EXCEPTION QUEUE === Exception Types: - Personal or business purpose unclear - Missing amount - Missing currency - Missing date - Vendor unclear - Unreadable image - Password-protected attachment - Unsupported file - Probable duplicate - Multiple transactions in one document - Refund relationship unclear - Attachment missing - Storage failed - Forwarding failed - Conflicting fields Exception Record: Source: [...] Reason: [...] Observed Data: [...] Missing Data: [...] Proposed Action: [...] Risk: [...] Reviewer: [...] Resolution: [...] === PARTIAL FAILURE HANDLING === Case: Forwarding succeeds but storage fails. Action: - retain Forwarded status - do not forward again automatically - retry storage only - record failure - notify the exception queue Case: Storage succeeds but forwarding fails. Action: - retain Stored status - retry forwarding only - do not create another file - record the failure Case: Processing log fails. Action: - preserve the successful forwarding and storage states - create a log-repair task - do not repeat previous successful actions === HISTORICAL BACKFILL PLAN === Period: January 1, 2026 to present Phase 1: Audit-only discovery Phase 2: Review search coverage and exclusions Phase 3: Classify business, personal, ambiguous, duplicate, refund, and exception records Phase 4: Approve folder and naming behavior Phase 5: Store approved historical documents Phase 6: Forward only according to finance-approved historical-forwarding policy Phase 7: Create backfill reconciliation report Do not flood the finance inbox with historical documents without explicit approval. === MONTHLY CLOSE WORKFLOW === 1. Confirm the month date range. 2. Review unresolved exceptions. 3. Review probable duplicates. 4. Review refunds and credits. 5. Check expected recurring subscriptions. 6. Identify missing receipts. 7. Review late arrivals. 8. Compare with supplied statement data. 9. Preserve separate currency totals. 10. Create month-end summary. 11. Mark unresolved items. 12. Obtain finance approval where required. === MONTH-END SUMMARY === Month: July 2026 Documents Discovered: [...] Business Confirmed: [...] Forwarded: [...] Stored: [...] Duplicates Skipped: [...] Refunds and Credits: [...] Late Arrivals: [...] Missing Expected Receipts: [...] Exceptions: [...] Failures: [...] Unresolved: [...] Currency Totals: Keep each currency separate unless an approved conversion policy exists. === AUDIT LOG FIELDS === - processing timestamp - source message identifier - sender - subject - email received date - attachment name - attachment identifier - vendor - transaction date - amount - currency - document type - expense category - business-purpose confidence - duplicate status - refund relationship - destination - forwarding status - storage status - exception status - reviewer - resolution - final status === ACCEPTANCE CRITERIA === High-Confidence Receipt: Given an approved vendor sends a valid business invoice, when the workflow processes the message, then the document is classified, checked for duplicates, stored in the transaction month, forwarded once to finance, and recorded in the audit log. Personal Purchase: Given a document contains a strong personal-purchase signal, when classification runs, then it is not forwarded automatically and is marked Personal or Needs Review. Duplicate: Given the same invoice was already processed, when a resent copy is discovered, then the workflow does not create another finance forward or stored file. Late Arrival: Given a July receipt arrives in August, when the transaction date is reliable, then the document is stored in the July folder and logged as a late arrival. Partial Failure: Given forwarding succeeds and storage fails, when retry processing runs, then only storage is retried. Refund: Given a refund matches an earlier expense, when processing completes, then the refund is linked to the original record and is not counted as a new expense. Missing Subscription Receipt: Given an expected monthly subscription receipt is not found, when monthly close runs, then it appears in the missing-receipt report without claiming that no charge occurred. === REGRESSION TEST MATRIX === Document Types: - Receipt - Invoice - Payment Confirmation - Subscription Charge - Refund - Credit Note - Cancellation Formats: - PDF - PNG - JPEG - HTML Message - Multiple Attachments - Multi-Page Document - Password-Protected File - Unsupported File Classifications: - Business - Personal - Ambiguous - Duplicate - Refund - Exception Dates: - Transaction Month - Late Arrival - Missing Date - Conflicting Dates - Year Boundary Failures: - Search Failure - Forward Failure - Storage Failure - Folder Creation Failure - Log Failure - Partial Failure - Authentication Failure Currencies: - USD - EUR - GBP - Multiple Currencies - Currency Missing === QA CHECKLIST === Email Search: - Verify date boundaries. - Verify approved vendors. - Verify exclusions. - Verify historical messages. - Verify already-processed filtering. Documents: - Verify attachment integrity. - Verify amount. - Verify currency. - Verify date. - Verify vendor. - Verify reference. - Verify document type. Classification: - Verify business signals. - Verify personal signals. - Verify ambiguity handling. - Verify category. - Verify confidence. Duplicates: - Verify source ID. - Verify attachment hash. - Verify reference. - Verify vendor-date-amount comparison. - Verify resent documents. Storage: - Verify year folder. - Verify month folder. - Verify document subfolder. - Verify file name. - Verify no overwrite. - Verify original document. Forwarding: - Verify recipient. - Verify subject. - Verify body. - Verify attachments. - Verify no unrelated content. - Verify single delivery. Exceptions: - Verify queue entry. - Verify reason. - Verify evidence. - Verify proposed action. - Verify final resolution. Privacy: - Verify masked data. - Verify approved recipients. - Verify no credentials. - Verify no unrelated personal correspondence. - Verify approved retention. === RISKS === Risk: Personal receipts are forwarded as business expenses. Mitigation: Use business-purpose rules and route ambiguous documents to review. Risk: Finance receives duplicate documents. Mitigation: Use idempotency records and multi-signal duplicate detection. Risk: A receipt is filed in the wrong month. Mitigation: Use transaction date first and record date uncertainty. Risk: A successful finance forward repeats after a storage failure. Mitigation: Track forwarding and storage independently. Risk: A refund is counted as a new expense. Mitigation: Classify refunds separately and link them to original records. Risk: Sensitive email content is exposed. Mitigation: Forward only the relevant document and concise approved metadata. Risk: Missing receipts are interpreted as missing charges. Mitigation: Report only that a document was not found; do not claim that no transaction occurred. === KNOWN LIMITATIONS === - Email and storage access require authorized integrations. - Document extraction may be incomplete. - Business purpose may require manual review. - Tax deductibility requires professional judgment. - Duplicate detection is probabilistic when identifiers are missing. - Currency conversion requires an approved policy. - Reconciliation depends on complete statement data. - Provider limits and authentication changes may interrupt scheduled processing. - The workflow does not replace bookkeeping, tax, audit, legal, privacy, or security review. === FINAL INTEGRITY NOTE === A receipt-like document is not automatically a valid or deductible business expense. All forwarding recipients, folder destinations, classification rules, accounting periods, tax treatment, retention requirements, refund relationships, and deletion decisions must be verified. The workflow is designed to improve receipt discovery, monthly organization, traceability, and finance-team efficiency. It does not guarantee accounting completeness, tax compliance, fraud detection, or successful reconciliation.