The age of scrolling endlessly through product pages may soon be over. The digital economy is shifting toward "agentic commerce"—a paradigm where AI agents act not just as search tools, but as active proxies that make decisions, negotiate, and transact on behalf of humans.

A critical question facing the next wave of digital transformation: How can we build responsible, trustworthy agentic AI that powers commerce at scale?

A panel titled “Agentic Commerce: Trust and Identity in the AI Economy” at the AI Impact Summit in New Delhi brought together experts from technology, finance, and academia to explore the challenges and opportunities of a new era where AI agents act as autonomous proxies for human users.

From e-commerce to agentic commerce

Moderator Dr. Pankaj Jalote framed the central dilemma: while AI agents can process vast data and automate tasks, human decision-making often relies on tacit knowledge — intuition and context that are not easily captured in data models. He said that until those nuances are better represented, agentic systems will face limitations in replacing human judgment.

Panelists generally agreed to the idea that agentic AI is not merely an incremental upgrade to existing e-commerce systems, but a fundamental shift in how transactions will be initiated and completed in the future.

Agents — defined as AI systems capable of interpreting preferences, making decisions, and completing tasks — are rapidly evolving from simple assistants to digital representatives capable of executing transactions on behalf of users.

“Agents are becoming our digital proxies,” said Citi's Prag Sharma, stressing that these systems will soon be trusted to act autonomously across financial, commercial, and personal domains.

Merchant demand and data evolution

PayPal's Dr. Prakhar Mehrotra mentioned a trend that merchants are actively seeking agentic capabilities. Unlike previous technological waves, where companies built first and searched for adoption later, businesses are now requesting AI agents to improve engagement and productivity.

The reason, he explained, lies in the shift from structured to unstructured data. Conversational agent interactions generate deeper preference insights than simple clicks. That feedback loop enables merchants to refine supply chains and personalise offerings more effectively.

Identity, authorisation, and provenance

A central theme of the discussion was trust — particularly how to ensure that agentic AI systems are secure, auditable, and aligned with user intent. Panelists agreed that without robust frameworks for identity and authorisation, the promise of agentic AI could falter.

Prag Sharma laid out a three-part trust architecture essential for responsible agentic systems: cryptographically verifiable identity, granular authorisation mechanisms, provenance and auditability — every action taken by an agent must be traceable and accountable, allowing users and systems to understand what decisions were made, and why.

These principles mirror traditional security frameworks but require reimagining for systems that learn and adapt.

When convenience meets risk in real-world use cases

Mastercard's Janet George shared a personal experience that highlighted both the convenience and risk of agentic systems. Faced with a last-minute need to fund her retirement account, she used an autonomous agent to move funds between accounts. The system repeatedly sought her consent for each step — a critical design choice that helped build confidence in its actions.

“That level of user interaction and consent is crucial,” George said. “It showed how trust can be engineered into autonomous workflows.”

Architecture, tokens, and standards

Beyond trust, the panel explored how structural elements like agent memory management, observability, and tokenisation can power scalable, secure systems. Tokens — digital representations of credentials, payments, or identity markers — were identified as key enablers of secure, traceable interactions between agents and services.

There is a definite need for global standards and governance frameworks to ensure interoperability and safety across borders. “AI knows no colour, no borders,” George noted, stressing that responsible agentic AI must be developed with equity, safety, and accountability at the forefront.

Inclusion in the AI economy

Panelists highlighted the potential for AI to aid decision-making across economic strata — from high-value financial products to everyday consumer choices — provided technologies are made accessible and affordable.

“If we can crack the cost of running agentic AI for low-value transactions, there’s an opportunity for universal benefit,” said Glance's Arvind Jayaprakash.

He also shared early performance data suggesting that agent-assisted commerce improves outcomes. When customers interact with AI agents before reaching a merchant platform, conversion rates can increase multiple times compared to traditional browsing, as indecision is resolved through guided dialogue.

The consensus was clear: the technology is ready, but the frameworks for trust and governance must now catch up.