Agentic AI in Payments: What's Real in 2026 as Mastercard, Visa Run Pilots Amid Regulatory Wait

Agentic AI Moves From Buzzword to Banking Pilots

Agentic AI has rapidly evolved from abstract research concept to active pilot programs in the payments industry. Mastercard and Visa have enlisted banks across Asia-Pacific to test agentic payment systems that allow AI agents to complete transactions on behalf of consumers3

3

. Singapore-based DBS became the first bank in the region to pilot Visa Intelligent Commerce, conducting real-world food and beverage transactions using credit and debit cards via secure, issuer-controlled flows3

3

. Separately, Mastercard worked with Westpac to complete New Zealand's first transaction using its Agent Pay framework, purchasing cinema tickets with full cardholder consent3

3

.

The Reality Behind Agentic Commerce Announcements

Despite prominent announcements, the distinction between marketing and actual deployment remains significant. Mastercard's "first fully authenticated agentic commerce transaction" in India was a controlled sandbox environment demonstration rather than a live commercial rollout, according to Gautam Aggarwal, Mastercard's India and South Asia President4

4

. While payments used real systems and real cards, merchants were connected through sandbox environments since no Indian platform is yet fully embedded into an AI agent ecosystem4

4

. The payment was real, but the commerce side remained controlled.

Regulatory Approval Holds the Key

Commercial rollout of AI-driven agentic commerce transaction capabilities depends entirely on regulatory clarity. In India, end-to-end AI-based commercial transactions are not currently allowed by the Reserve Bank of India, though fintech companies are building infrastructure in preparation2

2

. Aggarwal indicated that if regulators treat AI-led transactions as extensions of already approved technologies like tokenisation and passkeys, deployment could happen within months. However, if fresh approvals are required, the timeline extends significantly4

4

. Payment aggregators like Razorpay and Cashfree, along with card networks and merchant processors including PayU and Pine Labs, are all working toward integration with platforms like ChatGPT and Claude2

2

.

Where Agentic AI Actually Works in Payments Today

The most mature applications of agentic AI in payments exist in fraud, risk, and account takeover detection systems that have operated for years. According to industry data, 87% of global financial institutions now deploy some form of AI or machine learning-driven fraud detection1

1

. Visa launched generative AI-powered fraud capabilities aimed at detecting specific attack patterns, while Stripe Radar evaluates transactions in real time and autonomously intervenes based on risk1

1

. Large payment service providers like Adyen, Stripe, Worldpay, and Checkout.com use dynamic transaction routing and authorization optimization as standard capabilities1

1

.

Human-in-the-Loop Systems Dominate Current Deployment

Most agentic payment systems in production today operate as human-in-the-loop systems rather than fully autonomous actors. These tools automate analysis, summarization, and drafting while humans remain responsible for final decisions and execution1

1

. Providers like Stripe and PayPal use generative AI to help merchants cope with rising chargeback volumes, summarizing evidence and drafting responses1

1

. This category represents where the steepest growth is expected, functioning as decision support and acceleration layers.

The Infrastructure Being Built for Future Adoption

Fintech companies are actively preparing infrastructure for widespread agentic commerce adoption. Currently, UPI Reserve Pay is the only payment mechanism through which an AI agent can undertake a transaction in India, according to Razorpay CEO Harshil Mathur2

2

. This mechanism allows consumers to set transaction limits for specific merchants, enabling subsequent transactions within that limit to flow without an OTP. Card networks are pushing tokenized payments via AI agents, where real card data remain masked and tokens can be stored with AI platforms to ensure transaction safety2

2

. Pine Labs announced integration with OpenAI's ChatGPT to enable agentic commerce and plans to run experiments with AI agents completing full transactions in Southeast Asia and Gulf countries2

2

.

What the Market Momentum Reveals

More than three-quarters of Singapore residents already use generative AI tools like chatbots in daily life, with eight in 10 consumers relying on AI assistance for online shopping3

3

. This consumer behavior shift drives payment companies to prepare infrastructure even as regulatory frameworks lag behind technological capability. Spending on agentic AI is projected to reach $155 billion by 20301

1

. However, in payments, most spending does not yet flow into systems that can move money autonomously. The technology remains rail-agnostic and can operate across cards or UPI, making it particularly relevant for markets like India where both payment methods have wide adoption4

4

. As one digital payment firm CEO noted, while adoption remains very low, people are searching for merchandise on AI-powered apps, making transaction completion within those apps the logical next step2

2

.