BT EXCLUSIVE: Can the soundbox become a bank branch?

On a recent grocery run, a vegetable seller asked me to wait after I’d made the payment. His soundbox hadn’t announced the payment confirmation. After a few minutes of checking the device, he gave up. I had to eventually show him the payment confirmation on my mobile phone.

From Delhi to Dindigul, from the fruit seller to high-street retailers, the soundbox—a small plastic box that confirms a payment made via UPI—has become ubiquitous in India’s shopping experience. A UPI payment isn’t complete until the soundbox says so.

“If the machine doesn’t announce the payment, I get worried,” says Hariom, the vegetable seller from Greater Noida. “Last week, I lost Rs 700 because the machine didn’t announce a payment and I couldn’t check the customer’s phone. Now I have to figure out what is wrong with the machine.”

Must read: PhonePe brings internet-free UPI payments to feature phones, expanding digital access across India

“It would have been so much better if we could check with the machine itself in case something went wrong, instead of checking on the app or going to the bank,” says Samar Bhati, a grocery shop owner. “If I could ask the machine whether a payment has settled, or how much I received that day, I wouldn’t have to open the app every time.”

A handful of companies are now trying to figure out if the box can do more than confirm a payment. One of them is ToneTag, a Bengaluru-based company, which thinks it could become a bank branch.

“ATM was a bank branch extension for consumers,” says Kumar Abhishek, ToneTag’s founder and CEO. “Now these boxes becoming conversational, it becomes a bank branch extension for MSMEs.”

In July, ToneTag unveiled eKosha, a voice-first AI banking assistant for merchants. It builds on ToneTag’s decade-old sound-wave protocol, which turns data into sound, so two devices can talk to each other. eKosha takes it a step further, turning voice into data and processing most of it on the local device instead of sending it to the cloud. ToneTag is targeting India’s 65-80 million MSMEs, hoping to turn the payment device into a banking interface.

Must read: UPI accounts for nearly half of global real-time payments: How India became the world leader

A merchant could ask the device about a settlement, request a cheque book, seek information about a loan or access other banking services, without opening an app, making a call or visiting a branch.

As merchants moved from cash to UPI, they became more dependent on the soundbox. Yet when a payment failed or a settlement lagged, the only support on offer was a toll-free number or a digital app. “An anxious merchant in front of whom a customer is standing, to resolve an issue, he has to chat in a digital app,” Abhishek says. “And the app is in English.”

UPI’s next frontier

It is that gap ToneTag is trying to address through eKosha, which works in 10 Indian languages. It runs on UPI, UPI 123Pay and central bank digital currency rails and ToneTag says it requires no new infrastructure investment.

ToneTag is not alone in betting big on the soundbox. Paytm launched its own AI soundbox in October 2025. It can talk in 12 languages and answer questions about collections, trends and performance, and even give news updates.

“Today we bring the perfect AI device for millions of businesses… An AI system, with power and training of their payment and business data,” Paytm’s founder and CEO Vijay Shekhar Sharma said at the launch, describing it as a "very smart and capable business partner on the counter for 100s of millions of small businesses”.

In March 2026, Razorpay partnered with Sarvam AI to build voice-first conversational commerce, letting shoppers discover, order and pay by speaking, with Swiggy as the first pilot partner.

The broader push into voice and AI is becoming the next frontier for UPI. Dilip Asbe, managing director and chief executive of the National Payments Corporation of India, has said that AI could help bring the next half-billion users into the digital-payments system. He has also pointed to AI’s potential role in fraud detection, credit distribution, multilingual interfaces and simpler onboarding. NPCI launched a voice assistant-based payment system in 2023, but it is still far from becoming a mainstream way to make payments.

Must read; UPI and MDR: Can selective charges sustain India’s digital payments ecosystem?

The wider adoption of voice technology will require merchants to trust it with financial transactions, which raises questions about authentication and voice cloning.

“Voice interfaces confirm intent, not identity,” says Amit Jaju, senior managing director–India at Ankura Consulting, who leads the risk advisory, cybersecurity, and data and AI practices in India. “A merchant’s spoken command to check a settlement or request a loan can be intercepted, replayed, or synthetically recreated without the merchant touching the device.”

Jaju identifies three key risks: ambient noise attacks in crowded shops where someone nearby mimics or replays a command, social engineering calls that trick a merchant into repeating phrases later stitched into a fraudulent instruction, and connectivity-level compromise where a soundbox’s network stack is hijacked to inject spoofed audio directly.

Navigating security risks

To safeguard against this, Jaju prescribes three layers of protection. First, device-bound authentication where the soundbox is a trusted hardware anchor tied to a secure element, so a command from elsewhere doesn’t work. Second, mandatory secondary confirmation for anything that moves money or changes account settings, a PIN tap or app approval, not voice alone. Third, behavioural monitoring that flags unusual command patterns, timing, or transaction sizes against a merchant’s own history.

“For high-value transactions, inferred context alone shouldn’t carry full trust, it needs to sit alongside something cryptographic, like a device-bound token, not replace it,” says Jaju.

ToneTag’s answer is to embed the merchant’s identity inside their voice. When a merchant first registers, their phone number is linked to their device and their voice. After that, every time they speak to the box, that number travels inside the sound of their words. There is no separate code.

For banks, the pitch is about scale and cost. The rollout itself is small: Abhishek puts the current number of conversational devices at 30,000 to 35,000, with larger deployments expected in the second half of the year.

Banks are already using soundboxes as part of their merchant offerings. Depending on the bank, the device may do little more than confirm payments, or sit alongside a broader suite of services such as settlement reports, loans, insurance and merchant support. But those services still largely sit within an app or web interface.

The question for ToneTag, then, is whether putting those services behind a voice interface can make the soundbox a meaningful standalone banking channel rather than simply a smarter payment terminal.