NPCI Aims to Reach 1 Billion Daily Transactions with AI

The National Payments Corporation of India (NPCI) has announced plans to reach a billion daily transactions using artificial intelligence (AI), with its CEO Dilip Asbe stating that AI will be central to reaching this goal. The NPCI's Unified Payment Interface (UPI) has already grown to over 750 million daily transactions, and the organization is looking to leverage AI to drive user growth, fraud prevention, and credit distribution.

Background and Context

The UPI market in India remains heavily concentrated, with PhonePe and Google Pay controlling over 80% of transaction share. However, NPCI has extended the deadline for third-party UPI apps to comply with a 30% market-share cap until December 31, 2026, giving newer entrants more time to compete. The concentration problem has lingered for years, but Asbe believes that AI can help make the business model work for new entrants.

NPCI has already moved ahead on the technology side, launching a payments-focused AI model called FIMI in 2023. FIMI is handling nearly one million UPI support-service transactions every month, and NPCI sees an opportunity to build small language models tailored to local languages and use cases. Asbe believes that Indian companies can create sharp, specific, and deterministic language models using the rich data set available in the ecosystem.

Why it Matters

The adoption of AI in finance is a significant development for the industry, as it has the potential to improve fraud control, merchant tools, and credit underwriting. NPCI's goal of reaching 1 billion daily transactions using AI is ambitious, but it could also lead to increased competition on the UPI network. Asbe believes that with robust regulations and a framework, India can adopt AI-powered finance while protecting users and mitigating risk.

The use of voice and multilingual solutions is seen as a means to simplify registration and access to payments, but full adoption requires higher accuracy of voice models. NPCI launched a voice assistant-based interactive system in 2023, but its widespread deployment depends on specific use cases. Asbe believes that AI can help make voice models more accurate and effective.

What Comes Next

NPCI's plans to reach a billion daily transactions using AI are likely to be closely watched by the industry. The organization will need to work with regulators, government agencies, and other stakeholders to develop a robust framework for AI-powered finance. Asbe believes that Indian companies can create small language models tailored to local languages and use cases, but this will require significant investment in infrastructure and talent.

The adoption of AI in finance is not limited to NPCI's plans. Other organizations, such as IBM India & South Asia, are also working on AI-powered finance solutions. Sandip Patel, managing director at IBM India & South Asia, believes that companies need to have the right data for specific use cases to deliver meaningful outcomes. He also emphasizes the importance of cost optimization and sustained return on investment (RoI) in AI adoption.

Key Facts

• NPCI's Unified Payment Interface (UPI) has grown to over 750 million daily transactions.
• The organization aims to reach a billion daily transactions using artificial intelligence (AI).
• NPCI has launched a payments-focused AI model called FIMI, which is handling nearly one million UPI support-service transactions every month.
• Indian companies can create small language models tailored to local languages and use cases using the rich data set available in the ecosystem.
• The adoption of AI in finance requires robust regulations and a framework to protect users and mitigate risk.

NPCI's plans to reach a billion daily transactions using AI are ambitious, but they have the potential to improve fraud control, merchant tools, and credit underwriting. The organization will need to work with regulators, government agencies, and other stakeholders to develop a robust framework for AI-powered finance. Asbe believes that Indian companies can create small language models tailored to local languages and use cases, but this will require significant investment in infrastructure and talent.