What is the significance of devices in fintech?
Ripunjai Gaur: At Paytm, after 15 years in fintech and payments, the most visible change has happened in the past three to four years. What was once a passive device at a merchant counter has become an active interface. As devices moved to the Cloud, they turned into points of interaction where data is captured, and transactions are initiated, while processing happens in the backend. This evolution will continue. The Paytm soundbox, for example, is now an AI-enabled device that goes beyond announcing payments. It helps merchants and customers solve everyday problems.
At a recent Global Fintech Fest (GFF) demonstration, a foreign tourist described the difficulty of communicating with shopkeepers while making payments. With AI-enabled devices, real-time language translation is only a starting point. Combined with Cloud intelligence, such devices can fundamentally change commerce, especially in a developing country like India.
How are devices evolving, and what role does ‘made in India’ play?
Sivaramakrishnan Iswaran: We are seeing a major shift in how technology shapes banking. The first transition was from desktop-based banking to the Cloud. The next was the connected Cloud, where systems became interoperable. That led to contextual or embedded banking. Instead of logging into a separate banking portal, users now initiate transactions within the software they already use. At Zoho, with our made in India enterprise resource planning products, banking transactions can be initiated directly from finance applications (apps). We worked with ICICI Bank in 2018 to launch connected banking, well before it became mainstream.
Once banking became connected, the focus moved to devices. With systems already in the Cloud, AI and IoT naturally followed. Unified Payments Interface (UPI) demonstrated the power of connected devices.
Do we need made-in-India tools?
Mayank Sharma: This question has two dimensions. One is what Google is building in India. Gemini is an example. India was among the first global launch markets, and Gemini supports nine Indian languages. In live demonstrations, users often switch between languages such as Kannada and Marathi mid-conversation. These capabilities are being built in India
Security is another focus area. Android includes features such as life-threat detection that flag malicious activity, including calls that ask users to share screenshots from financial apps. Many of these features were developed by engineering teams in India. Google has also invested in an AI hub in Vizag.
From a device perspective, mobile phones are no longer just delivery channels. They are full computing platforms. Banks and fintech companies are increasingly using on-device AI. Open-source models allow multimodal interactions without a Cloud connection, while premium devices support built-in features, such as summarisation and image descriptions. Globally, banks are embedding these capabilities directly into retail banking apps. The device is becoming an AI compute engine, not just a delivery engine.
How are devices, technology, and lending converging?
L Guru Raghavendran: The core purpose has not changed: customers still want to take loans, open accounts, and invest. What has changed is the medium. These actions now happen through devices, supported by AI. Backend banking systems continue to process transactions, but branch visits have become rare, limited mostly to exceptions such as signature mismatches or loan closures.
For this model to scale, backend systems must be designed accordingly. API-first, microservices-based architectures are essential. Fraud has become more complex. Today, more than 450 parameters can be evaluated on a device to assess authenticity, compliance, and risk. As lending scales up, maintaining quality growth is critical.
How are risks and fraud being handled?
Gaur: Security remains central to banking and fintech, extending beyond financial loss to data protection, especially as credentials move to the Cloud. While data leaks still occur, their frequency has declined. Earlier, fraud prevention relied heavily on manual rules — blocking high-risk UPI IDs, accounts, or cards. Today, real-time computing and advanced detection mechanisms are used. Device-based controls such as UPI device binding have significantly reduced fraud. If a SIM changes or a one-time password is compromised, transactions are blocked automatically. With AI-enabled devices and backend analytics, user profiling has improved, even as fraud patterns become more sophisticated.
How is risk mitigation evolving?
Iswaran: AI engines are now widely used across the industry. When Zoho entered payments, fraud detection was a priority. Manual checks have been replaced by AI systems that correlate data to identify fraud clusters, such as shared email IDs, linked accounts, or unusual fund flows. Recently, a large fraud cluster was identified using these techniques. Coordinated action across devices, operating systems, technology providers, and banks is critical. Early detection and collective response can reduce fraud, and the savings can eventually be passed on to customers.
Are regulatory guardrails adequate?
Raghavendran: There are two aspects here: anti-money laundering (AML) and fraud. AML is well-regulated. Banks and non-banking finance companies are required to submit regular reports, and AI models are increasingly used as regulatory technology to support compliance.
Fraud is more fragmented. Credit card fraud, current account savings account fraud, lending fraud, and payment fraud all differ. Banks often rely on multiple, siloed systems to manage these risks, making it difficult to obtain a single customer view across products. Greater regulatory clarity on fraud would help, even as governments are beginning to address it.
How is fraud being managed at the platform level?
Sharma: Fraud involves both the user being targeted and the actor behind it. Internal threats must also be considered. In India, banking employs nearly 2 million people, and the banking, financial services, and insurance sector touches hundreds of millions. On the user side, platform-level tools play a major role. Android uses on-device AI for scam detection in messages. Models such as Gemini Nano run directly on premium devices, while open-source models like Gemma support multimodal analysis without relying on the Cloud. These models can analyse messages and app behaviour on the device, helping counter common fraud vectors such as malicious apps, messages, and URLs.
On the enterprise side, devices used by banking employees are typically managed. Messages on official devices can be archived to meet regulatory requirements, and security signals can be collected even on partially managed devices. This ensures that financial applications are accessed only from secure environments.
 "(L-R) Ripunjai Gaur, COO, Offline Payments, Paytm; Mayank Sharma, SE Asia Regional Head, Android Enterprise Partnerships, Google; Sivaramakrishnan Iswaran, CEO, Zoho Payment Technologies and L Guru Raghavendran, Senior V-P & Product Head, Azentio So")