Resolve AI

ResolveAI checks seller proof before submission, reconciling POs, invoices, and manifests with AI to clear PayPal holds instantly.

Comment

Inspiration

Every growing e-commerce merchant dreads the sudden notification: "Your funds are on hold." When a seller's business takes off or seasonal volume spikes, PayPal's risk engine places reserves or holds on payouts to protect against fulfillment fraud.

For honest merchants, this creates a cash-flow crisis. Today’s dispute and hold resolution process is painful: merchants blindly upload disorganized PDFs, wait 3 to 7 business days, and often receive generic rejections because a single line item didn't match. Meanwhile, human PayPal risk analysts spend hours manually cross-checking invoices, purchase orders, and shipping manifests. We built ResolveAI to turn this week-long friction into a 3-second automated verification.

What it does

PayPal ResolveAI is an intelligent evidence reconciliation platform that checks seller proof before submission:

  1. Automated Document Triage & Client-Side Redaction: Merchants upload invoices, purchase orders, and carrier manifests. Before any data leaves the browser, sensitive PII (credit cards, SSNs, phone numbers) is redacted and documents are cryptographically fingerprinted using SHA-256 hashes.
  2. Three-Way Evidence Reconciliation: The engine cross-references the three core pillars of commercial proof:
    • Wholesale Purchase Order: Validates buyer identity, order quantities, and agreed ship-to address.
    • Supplier Invoice: Verifies unit pricing, billed quantities, and physical goods classification.
    • Carrier Delivery Manifest / Proof of Delivery: Confirms carrier tracking, destination match, weight consistency, and recipient signatures for orders exceeding PayPal's $750 threshold.
  3. Real-Time Preflight Checklist: Flags quantity conflicts, mismatched shipping addresses, or missing delivery signatures in red with plain-language corrective instructions before submission.
  4. Fast-Track Resolution & Cryptographic Receipt: Once all preflight checks pass and PayPal Seller Protection criteria are met, risk analysts can release reserves in one click (or trigger simulated automated settlement), generating an immutable, time-stamped audit receipt.

How we built it

  • Full-Stack Application: Built with React 19, TypeScript, TanStack Start, and Tailwind CSS.
  • Client-Side Privacy Engine: Implemented local regex-based PII redaction and Web Crypto SHA-256 fingerprinting to protect consumer data before processing.
  • Preflight & Seller Protection Rules Engine: Built custom reconciliation logic mapping directly to PayPal's Seller Protection guidelines (checking delivery proof, tracking verification, address uniformity, and signature requirements for transactions over $750).
  • Simulated PayPal Event Stream: Modeled an event-driven architecture tracking real-time PayPal lifecycle events (PAYMENT.CAPTURE.COMPLETED, RISK.RESERVE.HELD, EVIDENCE.SUBMITTED, HOLD.RELEASED).
  • Interactive Multi-Role Demo: Designed a dual-view portal allowing judges to experience the flow as both the merchant sending proof and the PayPal risk analyst reviewing the case.

Challenges we ran into

  • Reconciling Unstructured Document Variations: Supplier invoices, wholesale POs, and carrier manifests come in completely different layouts, measurement units (kg vs. lbs), and naming conventions. Building a normalization layer that accurately catches discrepancies (like 100 units on a PO vs. 500 units on an invoice) without false alarms required meticulous parsing rules.
  • Balancing Privacy with Proof: Risk review requires proof of legitimate fulfillment, but merchants shouldn't expose sensitive consumer payment card numbers or personal identifying data. We solved this by enforcing client-side redaction before ingestion.
  • Modeling PayPal's High-Value Seller Protection Rules: Accurately reflecting PayPal's risk threshold—such as requiring signed proof of delivery specifically when order amounts exceed $750—to ensure simulated holds mirror real-world compliance standards.

Accomplishments that we're proud of

  • Created a seamless, zero-friction preflight checklist that reduces evidence rejection rates from days of back-and-forth to instant self-correction.
  • Designed three distinct, one-click test scenarios (Clean Wholesale Spike, Quantity Mismatch, Missing Delivery Signature) so anyone can immediately see the reconciliation engine catch subtle discrepancies.
  • Generated verifiable audit trails featuring cryptographic document hashes, exact timestamps, and structured decision metadata.

What we learned

  • How much merchant churn and support overhead can be eliminated simply by providing real-time pre-submission feedback rather than post-rejection explanations.
  • The depth and precision of PayPal's Seller Protection criteria and how automating evidence parity empowers honest sellers while keeping risk barriers high against fraud.

What's next for ResolveAI

  • Live PayPal Resolution Center API Integration: Connecting directly into PayPal's Disputes and Case Management REST APIs to ingest active merchant holds and auto-attach verified evidence packages.
  • Carrier API Webhooks: Direct integrations with FedEx, UPS, and DHL to auto-pull signed Proof of Delivery (POD) directly from tracking numbers.
  • Multi-Modal Document Vision: Using multi-modal AI models to extract handwritten delivery signatures, bill of lading stamps, and complex multi-page invoice tables.

Built With

Share this project:

Updates

Submission history