At the AI Impact Summit 2026, the session ‘Agentic Commerce: Trust, Tokens and ‘Know Your Agent’ for the AI Economy’ was moderated by Dr Pankaj Jalote, Director of Indraprastha Institute of Information Technology. The panel featured Janet George, Executive Vice President for AI at Mastercard; Arvind Jaiprakash, Senior Vice President at Glance; Dr Prakhar Mehrotra, Senior Vice President and Global Head of AI at PayPal; and Dr Prag Sharma, Director of Emerging Technologies at Citi.
The discussion explored how agentic AI systems are transforming digital commerce, enabling trusted transactions, secure tokenized value exchange, human oversight, and inclusive participation, while also examining the frameworks and governance needed to deploy these systems safely and responsibly at scale.
The session was held at Bharat Mandapam, New Delhi, and forms part of the AI Impact Summit 2026, taking place from 16 to 20 February, which brings together academics, industry experts, and practitioners to explore practical applications of AI across commerce, finance, governance, and other sectors.
At the start of the discussion, Jalote highlighted the challenge of capturing tacit human knowledge for AI systems, which is essential for effective agentic commerce. Sharma explained the limitations of relying solely on structured data, “It is this tacit knowledge that we will eventually need to transfer to our models. During these conversations, we repeatedly come back to the data as the make or break gaps. We don’t know what a user likes unless it is explicitly captured, and this will be a critical challenge in the long term.”
Furthermore, George emphasised that existing consumer data provides a starting point for AI agents to understand user preferences, “We actually know a little bit of what users like through the loyalty programmes and transaction records. Agents can understand preferences and constraints and build from there.”
On the same note, Jalote stressed the importance of human oversight, particularly in high-stakes decision-making, “Even in areas where humans need to make decisions, agents can provide support, leveraging existing data to enhance decision-making while maintaining human oversight.”
Sharma noted that every consumer interaction generates actionable insights, enhancing agentic decision-making, “Every time a user interacts with their device, they generate data. This data can make or break agentic decision-making. While more is needed, we already know a significant amount about user behaviour.”
George illustrated the trust dimension of agentic commerce with a personal anecdote,“On December 29, I realised I had not funded my 401k. The custodian required an agent to execute the transaction. The agent asked for consent three times for the exact amount and accounts involved. It saved me an hour at the bank and completed the transaction correctly. The agent was de-instantiated afterwards. Trusted agentic interactions like this can provide significant convenience.”
Furthermore, Mehrotra described how merchants are generating unprecedented unstructured data through agentic commerce, providing insights into consumer behaviour, “The first interaction with an agent creates unstructured data that has never existed before. Merchants can now understand consumer preferences directly, from cappuccino choices to product demand. Supporting this infrastructure is critical to scaling agentic commerce.”
Similarly, Jaiprakash highlighted the impact on engagement and conversion rates, illustrating advantages for merchants“We have observed a seven to nine times increase in engagement when commerce is conducted via agents. Iterative interactions help consumers make decisions, resulting in a higher-quality user base for merchants.”
In addition, George reflected on how AI is reshaping the role of merchants and storefronts, “Traditional storefronts are no longer the sole points of transaction. The role of a physical or digital community is about intent and shared experiences. For example, golf enthusiasts seek a golf-centred environment to connect. Transactions now happen worldwide, and merchant strategy must reflect that shift.”
Sharma then outlined the technical requirements for secure agentic commerce, including identity verification, authorisation and provenance, “Generative AI models must act on behalf of users with trust. Digital proxies require cryptographically proven identities, granular authorisation, and a chain of custody. Achieving these three ensures accountability and enables redress mechanisms.”
On the same note, Mehrotra emphasised security considerations for different interaction modalities, “Whether interactions are voice or text-based, authenticity must be verified. Systems must prevent sensitive information from inadvertently being exposed to AI models.”
Jalote also highlighted the critical role of human-in-the-loop governance, ensuring visibility and oversight, “Human-in-the-loop mechanisms are essential to ensure that agents act as genuine proxies. They provide visibility and control to prevent unintended outcomes and ensure users remain informed of high-stakes decisions.”
Turning to inclusion and global access, Jaiprakash discussed how agentic AI can benefit consumers across all socioeconomic strata, “Decision-making complexity exists across society. The cost of technology determines whether low-value transactions are supported. If optimised, agentic AI can reach every consumer.”
Moreover, Mehrotra elaborated on the global reach enabled by agentic commerce, “Agents allow any merchant to access a worldwide audience. Language is no longer a barrier, and global transactions can be executed overnight. This is a paradigm shift, not incremental change.”
George concluded on the importance of governance and responsible AI practices, “AI transcends borders, bringing equality between large and small merchants. Global standards are essential to ensure transactions are safe. Inclusion is critical, but responsibility must guide implementation.”
Finally, Sharma emphasised the role of data-driven personalisation in enhancing commerce outcomes, “Algorithms will adapt to align merchant offerings with consumer preferences. As more data is generated, niche products and services can be accurately matched with the right consumers.”
Throughout the discussion, Jalote noted that trusted identities, human judgement and robust infrastructure are key to scaling agentic commerce. The session at Bharat Mandapam demonstrated that technical rigour, governance and inclusion will determine the future of the AI-driven commerce economy globally.
