After revolutionising retail payments by making transactions invisible at the point of sale, India is now piloting a bold innovation that could make the very act of buying vanish from traditional user interfaces altogether. On October 9th, the National Payments Corporation of India (NPCI), in collaboration with fintech giant Razorpay, OpenAI, and a consortium of partner banks, launched a pilot integrating the country’s ubiquitous Unified Payments Interface (UPI) directly into ChatGPT. This integration enables an AI-powered agent within ChatGPT to place user-authorised orders with select merchants initially including BigBasket, one of India’s largest online grocery platforms — without the user ever leaving the conversational thread. The significance of this pilot transcends mere convenience; it signals a potential paradigm shift in digital commerce where the boundaries between conversation and transaction blur to near invisibility.
India’s UPI ecosystem is already the backbone of an immense volume of digital payments, processing over 20 billion transactions monthly. The scale and reliability of these rails are not in question. What matters now is building agentic commerce — where AI agents act autonomously on behalf of users — that is secure, permissioned, and seamless enough to become “boring” in the best sense: trustworthy, frictionless, and ubiquitous. The pilot outlines several safeguards designed to ensure the integrity of the system. These include explicit user consent at every step, bank-verified credentials to authenticate identities, a limited roster of participating merchants to contain risk and complexity, and fallback protocols that hand off transactions to human operators if the AI encounters uncertainty or ambiguity. By anchoring AI-driven commerce firmly within existing regulatory and compliance frameworks, the pilot aims to mitigate the risks that have dogged broader AI deployments — from fraud and misuse to privacy violations.
If successfully scaled, agentic checkout could reshape the contours of digital shopping. For consumers, it promises fewer taps, less time spent navigating apps or websites, and a more natural, conversational interface for commerce. Merchants stand to benefit from new conversion funnels that harness intent signals embedded in dialogue rather than relying on traditional advertising or website visits. For banks and fintechs, the availability of richer risk data — including not just payment events but the underlying intent and context — could enhance fraud detection and credit underwriting. Unlike conventional digital payments that capture a snapshot of a transaction, agentic commerce provides a narrative thread, opening new avenues for smarter, more nuanced financial services.
Pilot live (Oct 9): NPCI + Razorpay + OpenAI integrate UPI into ChatGPT.
20B+/mo rail: UPI’s scale + consented AI = fewer taps, safer funnels.
Yet the technology does not come without risks. Consent fatigue, where users grow weary of repeatedly authorising AI actions, looms as a significant challenge. There is also the spectre of prompt abuse, where malicious actors might manipulate AI to initiate unintended transactions, and the risk of merchant impersonation, potentially duping consumers into paying fraudulent vendors. The pilot addresses these threats through a combination of rate limiting to curb abuse, robust multi-factor authentication to confirm user identity, and transparent, user-friendly interfaces that clearly communicate transaction details and permissions. These measures are vital not only to protect users but also to build the trust necessary for agentic commerce to gain widespread adoption.
India’s competitive advantage in this nascent field is not just the ubiquity and maturity of UPI’s digital payment infrastructure but also the country’s ingrained digital public infrastructure ethos. Unlike many other markets where payment and commerce platforms operate as isolated silos, India’s digital ecosystem is designed to foster interoperability and open dialogue between platforms. This culture of “platforms talking to each other” underpins the feasibility of embedding payments into AI conversations and could catalyse a wave of innovation in conversational commerce. As NPCI’s pilot scales, “conversational commerce” could become a literal reality no banners, no browser tabs, just user intent seamlessly transformed into fulfilment.
The timing of India’s experiment is notable. Globally, the rise of large language models and generative AI has spurred all manner of speculation about the future of commerce and payments. While many Western fintechs and tech companies are exploring voice assistants and chatbots for customer service or sales support, India’s approach is more ambitious, leveraging its mature payment infrastructure and regulatory frameworks to pilot true end-to-end AI-driven transactions. The involvement of OpenAI, a leader in conversational AI, coupled with Razorpay’s deep fintech expertise, signals a collaboration that blends cutting-edge AI research with pragmatic payment innovation.
From a regulatory perspective, India’s approach exemplifies a cautious yet proactive stance. The Reserve Bank of India (RBI) and NPCI have historically been supportive of innovations that expand financial inclusion while maintaining strong consumer protections. The AI-powered agentic commerce pilot fits this ethos perfectly: it pushes boundaries without discarding the hard-earned trust embedded in India’s digital payments ecosystem. Moreover, by involving banks directly as partners rather than merely as processors, the pilot ensures that financial institutions remain central to risk management and compliance, rather than being sidelined by third-party intermediaries.
The potential economic impact of agentic checkout is substantial. India’s e-commerce market, valued at over $100 billion and growing rapidly, faces persistent challenges related to user experience, cart abandonment, and payment friction. By enabling consumers to complete purchases through natural conversations, agentic commerce could reduce friction and increase conversion rates, especially among first-time or less tech-savvy users. For merchants, the ability to tap into AI-generated intent signals offers new avenues for personalised marketing and demand forecasting. Fintech firms and payment providers can also leverage richer behavioural data to refine credit products and fraud models. Collectively, these improvements could accelerate India’s ongoing digital commerce revolution.
However, widespread adoption hinges on more than technology and regulatory safeguards. Consumer trust and acceptance remain paramount. Early feedback from pilot users will be critical in shaping interface design, consent mechanisms, and fallback protocols. Clear communication about the AI’s capabilities and limitations will be indispensable to prevent misunderstandings and build confidence. Additionally, given India’s linguistic diversity, the AI must support multiple languages and dialects to be truly inclusive. There are also questions about how the system will handle disputes, refunds, and customer service escalation areas where human intervention will remain essential for the foreseeable future.
Privacy considerations loom large as well. Integrating AI agents with payment systems entails the collection and processing of sensitive personal and financial data. While India’s data protection regime is evolving, the pilot must ensure compliance with all applicable laws and industry best practices. Transparency around data usage, robust encryption, and user control over data sharing will be critical to prevent misuse and uphold consumer rights.
Beyond India, the pilot offers a glimpse into the future of AI-enabled commerce globally. Developed economies with fragmented payment systems and less interoperable digital infrastructures may struggle to replicate India’s seamless integration. However, the principles demonstrated explicit consent, strong authentication, limited merchant scope, and human fallback could inform safe and scalable models elsewhere. Moreover, as AI capabilities advance and regulatory frameworks mature, agentic commerce could become a standard feature of digital payments worldwide.
Industry observers note that India’s “agentic checkout” pilot aligns with broader trends in the convergence of AI, fintech, and e-commerce. Companies like Amazon and Google have long experimented with voice-activated shopping assistants, but these have often been hampered by partial integrations with payment systems and limited merchant ecosystems. India’s UPI-based approach, by contrast, leverages a unified payment rail that is open, interoperable, and backed by a strong regulatory framework. This gives it a potential first-mover advantage in deploying AI-driven commerce at scale.
In conclusion, India’s pilot of agentic checkout inside ChatGPT represents a pioneering step towards a future where commerce is seamlessly embedded into everyday conversations. By combining the scale and robustness of UPI with the intelligence and natural language capabilities of OpenAI’s models, the experiment presages a new era of conversational commerce that could reduce friction, enhance security, and unlock new economic opportunities. The success of this pilot will depend on careful attention to user experience, trust, privacy, and regulatory compliance. But if these challenges are met, India may well set a template for how AI and digital payments can come together to transform the way people shop.