What happened
Dilip Asbe, Managing Director and CEO of NPCI — the organization that operates UPI, India's unified payments interface running at 750 million daily transactions — told TechCrunch at Mumbai Tech Week on June 27 that AI will drive the system toward over 1 billion daily transactions.
Asbe named three specific AI applications. First, user acquisition: voice and multilingual AI tools to bring in users who haven't yet joined digital payments. Second, fraud detection: AI for identifying money mule operations and protecting existing users. Third, credit distribution: AI to assess creditworthiness from digital transaction footprints for users and merchants who lack traditional credit histories.
NPCI has already shipped AI in this domain. FIMI, its AI language model for payment disputes, now serves over one million users for mandate cancellations and issue resolution. Asbe also highlighted India's opportunity to build small language models tailored to payments: 'models will differentiate from each other based on the datasets that are made available to them.'
AI must be used to provide credit to all the users and merchants who have digital footprints and to make onboarding simpler through voice technologies.
Why it matters
The payments infrastructure running at NPCI's scale is applying AI to the same three-phase problem every subscription business faces: bringing customers into the billing relationship, protecting revenue from fraud and failures, and recovering it intelligently when something goes wrong. The difference is scale. A Stripe-native subscription business can apply exactly the same logic to its own billing data.
Asbe's fraud detection application has an underappreciated parallel in subscription billing. Failed payment charges are not all recoverable involuntary churn. Some are cardholders who dispute legitimate subscription fees. AI-based pattern recognition at the transaction level can distinguish between a genuine card failure and a behavioral pattern suggesting dispute risk — allowing subscription businesses to route those cases differently rather than treating every failure as the same recovery opportunity.
What this means for subscription businesses
The AI payment infrastructure is built — you just need to use it
What NPCI is building at national scale — intelligent failure classification, fraud pattern detection, dispute routing — is available to Stripe-native businesses through smart retry logic, decline-code-aware sequences, and risk-segmented dunning.
Not all failed charges deserve the same recovery sequence
Asbe's fraud detection focus applies here. A subscriber who disputes charges after recovering them is a different case from one whose card simply expired. Your recovery sequences should treat them differently.
Payment history is a data asset, not just a transaction log
Asbe's credit access application treats digital transaction history as a signal for creditworthiness. Your billing history — who recovers after failures, who churns, who disputes — is the same kind of structured signal for customer quality scoring.
The bottom line
The world's largest real-time payments system is using AI to push past a billion daily transactions by solving exactly the three problems subscription businesses face: bringing in users, detecting failures intelligently, and recovering revenue. The tools to apply that same framework to Stripe billing already exist. The NPCI story is a reminder that the organizations investing in AI-native payment infrastructure are the ones that will outperform the ones that treat billing as a passive process.
Sources