How agentic AI could reshape the way we make digital payments

Artificial intelligence (AI) is the hot topic right now, with industries such as IT already seeing a structural change with the adoption of new technologies. The payments industry is also integrating AI into its services, particularly agentic AI.

Conventional AI has been integrated into the payments ecosystem for quite some time, helping reduce friction in the payments process and enhancing digital safety by making the backend process more reliable, and screening consumers from potentially fraudulent transactions. However, agentic AI might take the industry towards agentic commerce, where an AI agent will take care of a consumer’s entire payment journey, making it frictionless.

The Global Fintech Fest (GFF) held in Mumbai last week saw many of the country’s top payments companies showcase AI integration into their products. The National Payments Corporation of India (NPCI) unveiled MyUPI — an AI-powered revamp capable of delegating pre-authorised payments, and automatic and seamless filing of payment disputes. It also launched a backend tool that acts as a “connective tissue” for AI interaction across the financial ecosystem, called Agentic Orchestration and Messaging (AtOM).

While agentic AI — a model that executes tasks for users without constant prompts, but within certain limits — could streamline the entire “procurement to payment” process for businesses, its adoption is still at a nascent stage. At present, agentic commerce happens to be its biggest use-case.

What is agentic commerce?

Agentic commerce is a digital trade model under which AI agents streamline the buying process by discovering, negotiating, and executing on behalf of customers. Notably, this does not mean bypassing the need for authorisation. It only cuts down steps involved in a customer’s e-commerce journey, including the payments phase.

For example, AI-powered tools can notify a customer of discounts on products that they have been viewing for months. If they decide to buy it, the AI agent on their payments platform can determine the ideal card or mode of payment — cutting down the need for multiple OTPs and authorisation requirements to just one-time authentication — for the transaction. Thus, the consumer will have oversight, and can change any of the components during the process. But they will save their time, and may find the best deal if they proceed with the whole process.

Amazon Pay’s smart wallet, demonstrated at the GFF last week, is one such example. “With agentic AI, I think that assistance for humans to get their job done in a very informed, intelligent manner, with a lot of control and trust, is kind of a large opportunity we see. And how that companion fits it into what we call a Smart Wallet,” said Girish Krishnan, Director of payments, rewards, and merchant services at Amazon Pay India.

Story continues below this ad

The Smart Wallet uses features such as smart recommendations, biometric authentication, and a tap-and-pay feature, and an AI agent taking care of each step to save the consumer’s time. It also has the ability to auto-execute small-ticket regular purchases within a particular limit. For instance, if a customer with the monthly limit of Rs 10,000 orders groceries, they won’t have to authenticate the process each time until the grocery amount for a month exceeds the threshold.

Such a feature is now being widely adopted by the payments industry. MyUPI has a similar feature. Other digital wallets provided by Samsung Pay, Google Pay, and PhonePe also offer similar features that enable pin-less small-ticket purchases. While not AI, Amazon Pay adds to this feature with its agentic commerce ecosystem.

Role of AI in B2B payments space

AI could also play a big role in the B2B payments space. “For merchants, it could predict settlement and cash-flow gaps, identify checkout failures, recommend the right acceptance mix, and automatically initiate refunds or customer recovery journeys. In B2B commerce, agents could reconcile invoices, match purchase orders, schedule payments, and optimise working capital,” according to Ranadurjay Talukdar, Partner and Payments Sector Leader at EY India.

EY estimates that agentic AI could automate around 80% of routine treasury coordination, and improve cash-forecast accuracy to nearly 90%, subject to strong data and governance foundations, added Talukdar.

Story continues below this ad

Agentic AI-driven payments solutions could also help smaller businesses specifically. “Bigger businesses have their own ERP (enterprise resource planning) software. But AI agents in the payments ecosystem can help the smaller SMEs (small and medium enterprises) due to their predictive modelling capabilities. Work is going on in that respect too,” an executive from a payments infrastructure company said.

BharatPe, for example, launched its own agentic AI solutions for merchants at the GFF, helping them understand merchant requirements and “take action across more than 60 live systems in real time.” Razorpay also showcased products that help agentic AI ease the checkout process for consumers through conversational models, enable cross-border payments, and enhance surveillance.

While AI integration is a promising proposition, adoption is at an early stage. “Adoption is meaningful but uneven. AI is well established in high-volume, data-rich areas such as fraud detection, risk scoring, customer support, and transaction analytics. However, many institutions remain at pilot stage because payment data is fragmented across banks, gateways, merchants, and legacy platforms,” noted Talukdar.

The dilemma: Who bears the cost of AI integration

Despite the promise, doubts remain about who will bear the AI integration cost. Training AI models takes up time and capital, and businesses will want to see a return on their investments. Both sides of the argument exist in the industry.

Story continues below this ad

One section feels that the current efficiency gaps are too small to get customers to pay additional amounts for AI services. “Today, the industry is paying for AI research and development. But obviously someday, we would like the customer to share that burden too. Indians are very price-sensitive, and it would take a big innovation for them to pay any additional cost. But today, we are mostly seeing AI (being) used to reduce some steps in the payments process, which I don’t think is a good enough efficiency at the moment,” said a Director at a small-sized payments services and infrastructure company.

Another section, however, remains optimistic. The view is that the efficiency gains, while currently in initial stages, will become too big for consumers to ignore. While the cost-benefit analysis of AI integration into the industry needs to play out, these services will achieve a “product-market fit” that could help the industry scale its business while driving up efficiency for consumers, said Amazon Pay India’s Krishnan.

“Over a period of time, I think we will hit that product-market fit. People say ‘Listen, this is too expensive. It does not have the ROI.’ This product-market fit in terms of smart recommendations, routing, etc., these are good ones for AI to really help us in,” he explained. Despite the current high costs, technology will be democratised over a period of time, which will reduce costs and drive new businesses.