Razorpay launches Vulcan AI model to improve payment routing and fraud decisions
Razorpay has launched Vulcan, a transformer-based payments foundation model trained on billions of masked historical transactions and thousands of payment signals. Its first applications span payment routing, international fraud, domestic chargebacks and checkout personalization, with an expected 8%–10% lift in payment success rates.
What happened
Razorpay launched Vulcan, a transformer-based AI foundation model for payments. Initially used for routing, international fraud, domestic chargebacks and
Key facts
- 8% to 10% payment success-rate improvement
- four initial use cases
- billions of historical payments
- thousands of payment signals
Why this matters
Razorpay’s AI-led payment stack raises the strategic value of acquiring or partnering with fraud-data, cross-border payments, and merchant personalization specialists that can deepen its model advantage.
What to watch
- Independently verified merchant cohorts showing whether the claimed 8%-10% payment-success lift persists after controlling for mix and seasonality.
- Changes in fraud-loss rates, chargeback ratios, false-decline rates and merchant disputes following deployment.
- Razorpay disclosures on Vulcan adoption, share of volume routed by the model, enterprise wins and cross-border penetration.
- New AI-routing or foundation-model announcements from Indian PSPs, banks, card networks and global processors.
- Regulatory guidance on AI-driven fraud, credit-like risk scoring, data masking, automated decisioning and payment-data localization.
- Issuer or network partnerships that expand real-time authorization, decline-reason or fraud-signal access.
- Offer Vulcan-powered routing as an opt-in merchant dashboard with approval-rate, cost, fraud and chargeback lift reporting.
- Use initial gains to bundle cross-border fraud, domestic chargeback automation and checkout personalization into higher-value enterprise contracts.
- Negotiate deeper issuer, bank and network data partnerships to improve approval prediction and recover more soft declines.
- Build merchant-configurable risk thresholds, model explanations and audit logs to support enterprise adoption and regulatory compliance.
- Competitors are likely to market AI routing guarantees, reduce take rates, pursue bank-data partnerships and emphasize transparent fraud governance.