Razorpay unveiled Vulcan, an AI model trained on 4 billion payments and 3 trillion data points. It promises 8-10% higher success rates and 8x fraud detection, but Razorpay keeps key data usage details proprietary. Merchants like Blinkit and redBus are already using its capabilities without clarity on how their payment data contributes.
How We Got Here
Razorpay developed Vulcan from the ground up, stemming from an earlier internal study of 1.5 million shoppers and 51,000 businesses. The model, built in partnership with Nvidia and AWS, represents a significant investment in proprietary AI amidst calls for data transparency in India's digital payments landscape.
The Numbers
- Vulcan processes transactions using around 3,000 signals, analysing patterns across merchants, payment instruments, and gateways.
- Early adopters like Blinkit, Bachatt, and redBus are already leveraging parts of Vulcan in live payments.
- Beyond success rates, Razorpay claims Vulcan identifies 5x more fraudulent transactions without increasing alert volume.
- The model also enables 40% more shoppers to see their preferred UPI app via Magic Checkout, driving 1-2 lakh additional purchases monthly.
- Razorpay developed Vulcan with proprietary architecture and training data, and is not charging merchants for its use.
What Happens Next
🇮🇳 Why This Matters for India
For India's 10 million small and medium enterprises (SMEs) in Tier-2 cities like Lucknow and Jaipur, who rely heavily on digital payments, clearer data usage policies are critical to trust new payment technologies.
The Take
Razorpay's opacity on data usage is a major red flag. True adoption for Vulcan will depend less on claimed success rates and more on transparent answers about merchant data ownership and sharing.
Source:
MediaNama ↗