India's UPI network is already moving close to 750 million payments a day. Dilip Asbe's AI pitch is not about novelty, it's about making the next 250 million daily transactions easier to start, safer to approve, and cheaper to support.
NPCI chief Dilip Asbe is putting artificial intelligence on top of one of the busiest payments systems in the world, and you should read that as a distribution move as much as a technology move. UPI processed 23.2 billion transactions in May 2026, worth Rs 29.90 lakh crore, according to figures reported by The Economic Times from NPCI data. That works out to 748 million transactions a day. The next round of growth won't come from telling urban smartphone users to scan another QR code. They already do.
The new fight is over the user who still finds payments confusing, the merchant who needs faster dispute handling, the customer who wants support in Tamil, Hindi, Bengali or another Indian language, and the app that can make all of this feel ordinary. Navbharat Times reported on Sunday that NPCI is preparing AI features for BHIM and looking at smaller Indian-language models to make payments easier across regions. If Asbe wants UPI to move toward 1 billion daily transactions, that is the right place to look.
Look at the gap. UPI has national scale, but BHIM does not own the consumer interface. The Economic Times reported in May that PhonePe and Google Pay command more than 80 percent of UPI payments, while Paytm accounts for another 10 percent. BHIM, developed by NPCI BHIM Services Ltd, has about 1 percent market share among UPI payment service providers, even after a sharp improvement in FY26. Monthly transactions on the app rose from 5.93 crore in April 2025 to 21.6 crore in March 2026, a 300 percent increase, according to another Economic Times report.
That is why AI inside BHIM matters. It gives NPCI a chance to compete on the interface without breaking the open rails that made UPI powerful in the first place.
Payments don't need a chatbot for the sake of having one. They need fewer failed journeys. A user who can't find the right button, doesn't understand a warning, mistypes a payee, forgets a PIN flow, or abandons a support ticket is not a theoretical problem at this scale. Multiply that by hundreds of millions of daily transactions and a small design issue becomes a national payments issue.
Indian-language models are a practical answer because India's payments market is not one language market. NPCI has already pushed UPI beyond the neat smartphone app story. BHIM has added biometric authentication for UPI payments up to Rs 5,000, and NBSL has been working to expand the app's reach to feature phones, The Economic Times reported in March and May. AI support in local languages fits that same direction. You don't get to 1 billion daily payments by making English-speaking users slightly happier. You get there by making the next user less nervous about pressing pay.
Fraud is the harder part, and frankly it is the part NPCI can't afford to dress up in vague AI language. A model that explains a suspicious collect request in plain language, flags unusual payment behavior, or routes a complaint faster has real value. A model that confidently gives the wrong instruction around money is a liability. UPI is not a shopping newsletter. It moves cash from bank accounts.
The Reserve Bank of India's December 2025 payments report, cited by The Economic Times, shows why the stakes are so high. UPI accounted for 85.5 percent of all payment transaction volumes in India in the second half of calendar 2025, but only 9.5 percent of total payment value. This is the rail for everyday retail life, not the rail for a few large transfers. Your tea stall payment and your grocery run sit beside hundreds of millions of similar low-value transactions every day.
BHIM needs more than cashback
BHIM's FY26 growth was helped by cashback offers, according to NBSL's own comments reported by The Economic Times. Cashback can wake up an app. It doesn't build a habit by itself. If PhonePe and Google Pay already sit on the home screens of India's most active UPI users, BHIM needs a reason to be opened that isn't just a discount.
AI could be that reason if it makes BHIM visibly easier for the user who has been left at the edge of digital payments. Imagine a support flow that understands a failed transaction in Marathi, a merchant onboarding flow that doesn't require a user to parse banking terms, or a voice-led payment journey that warns you before you approve the wrong request. Those examples sound plain because they are plain. Good infrastructure usually is.
NPCI has already tested the connection between AI and payments with larger partners. In October 2025, The Economic Times reported that NPCI and Razorpay had joined OpenAI to pilot agentic transactions through ChatGPT, with the idea of allowing conversational retail payments through UPI. That was the flashy version. The BHIM work may prove more important because it sits closer to the public rail and the mass user.
There is a risk here for every fintech founder watching India. If NPCI makes AI-native payments work inside BHIM, the strongest interface may not belong forever to the app with the largest current UPI share. If it doesn't, PhonePe and Google Pay keep the consumer relationship while NPCI keeps carrying the volume underneath.
UPI's next milestone is not mysterious. From 748 million daily transactions in May to 1 billion, the system needs roughly another 252 million payments a day. The hard part is making those payments come from people and places where the current interface still feels like a barrier. Asbe's AI plan is current because it points directly at that barrier. Now NPCI has to prove the models can do more than talk.
Also read: GM wants Cruise's failure to pay off in your driveway • AI SEO Is How Startups Get Found on Google and ChatGPT • The BIS says the AI boom has become a financial stability risk
