The Rise of Agentic Commerce

For decades, online shopping followed a familiar ritual: search, scroll, compare, add to cart, checkout. Interfaces evolved, payments became faster, and recommendations grew more sophisticated. Yet, the responsibility for decision-making always rested with the consumer.

That dynamic is now beginning to change. As AI agents step in to handle shopping tasks, e-commerce is moving towards automated decision-making.

This shift, known as agentic commerce, represents a deeper transformation than online retail has seen since the rise of digital marketplaces. Unlike traditional search and recommendation engines, AI agents can independently manage multi-step tasks such as product discovery, cart creation, checkout, and payments—all with user consent.

The economic potential is significant. McKinsey research estimates that agentic commerce could generate up to $1 trillion in orchestrated revenue in the US B2C retail market by 2030, with global revenue projections spanning $3 trillion to $5 trillion.

Consumer behaviour is already moving in this direction. As per Deloitte’s Connected Consumer Survey, more than 50% of consumers are now experimenting with or regularly using generative AI tools, signalling that AI-assisted discovery is moving rapidly into the mainstream. Younger consumers, in particular, are no longer starting their journeys on brand websites. They rely on platform-native search, recommendations, and increasingly, conversational interfaces. By 2030, analysts estimate that AI agents could handle up to 25% of global e-commerce transactions.

Companies have already started building for this future. Google recently launched Universal Commerce Protocol (UCP), an open standard designed to allow AI agents, retailers, and payment systems to work together across the entire shopping journey, from product discovery to checkout and post-purchase support.

UCP has been co-developed with retailers and platforms, including Shopify, Etsy, Wayfair, Target, and Walmart. It is also endorsed by companies such as Flipkart, Visa, Mastercard, and Stripe.

Balu Chaturvedula, SVP and country head, Walmart Global Tech, tells AIM that the company is introducing a bold vision: a company-wide framework of four intelligent super agents to reinvent experiences for customers, associates, partners, and developers.

“Our recent partnership with Google to bring the Walmart experience directly into Gemini, along with our collaboration with OpenAI through ChatGPT, are important steps toward creating agent-led commerce solutions that don’t just meet people where they shop, but anticipate their needs before they do,” he says.

Today, consumers typically make purchases directly from brand websites or specific merchants. With UCP, however, buying behaviour could shift towards completing transactions directly within Google Search. The search giant said UCP will soon power a new checkout feature on select product listings in AI Mode in Search and the Gemini app, allowing shoppers to purchase from eligible US retailers while they are still researching products on Google.

The feature is built with security at its core, allowing users to pay via Google Pay using saved payment and shipping details from Google Wallet, with PayPal support to be added soon.

Vijeth Pandit, CPO at Setu and Pine Labs, says that identity and consent will emerge as the real moat in AI-led commerce, as transactions increasingly happen within interfaces rather than on standalone websites. 

In such a model, the enduring customer relationship is not the UI itself but the underlying consent and identity layer—including saved preferences, mandates, wallets, subscriptions, and post-purchase service. Brands that control these rails, he says, will retain customers even as interfaces change.

He also outlines two possible futures for digital commerce: a closed model and an open one. In a closed system, a handful of dominant interfaces act as gatekeepers, forcing brands to pay “rent” for visibility. By contrast, an open model is built on interoperable rails that allow brands to plug in directly and compete on quality, trust, and fulfilment.

Last year, Razorpay partnered with the National Payments Corporation of India (NPCI) and OpenAI to launch Agentic Payments on ChatGPT. Khilan Haria, chief product officer at Razorpay, tells AIM that the feature enables instant payments.

“Once you choose, the transaction completes immediately. For this to work seamlessly, payment infrastructure must be invisible and instant, which is exactly what we’re building at Razorpay,” he says.

Haria adds that the company recently demonstrated the capability at the Global Fintech Fest (GFF), showing how AI agents can securely initiate and complete transactions within a single conversational flow.

Shopify, which provides commerce infrastructure to millions of online merchants, has also weighed in on how agentic systems could reshape both transactions and product discovery.

Shopify president, Harley Finkelstein, in an interview with CNBC, says the system is built to be highly flexible, allowing commerce to adapt across discounts, loyalty programmes, subscriptions, and delivery options, all of which can be integrated into AI agents. He adds that these systems will also manage brand discovery by aligning options with users’ needs and preferences.

This development is exciting, especially as the future of advertising may evolve, with brands relying on agents to improve product discoverability.

“Unlike traditional models, where shoppers entered static search terms, AI will help consumers refine preferences through conversational searches. Instead of browsing categories or keyword search bars, consumers will express needs in natural language, and AI agents will curate and rank products based on relevance and the precise need, not ad spend or SEO,”  Praveen Govindu, partner, Deloitte India, tells AIM. 

He adds that traditional brand websites and marketplaces will increasingly serve as data sources, while front-end engagement will slowly shift more toward AI-led conversations. 

For example, a user could ask an AI agent for furniture recommendations. After understanding the user’s preferences, it could suggest a specific set that best meets their needs. “You can then click the recommendation to immediately see the cost, including shipping and tax, with a button to pay via payment gateway. And it all happens without leaving the chatbot.”

In an agent-driven ecosystem, competitive advantage shifts away from how compelling a storefront looks and towards how reliably systems expose inventory, pricing, availability, and fulfilment through clean, real-time interfaces. Commerce, in this model, becomes less about where users shop and more about how systems connect behind the scenes.

Haria compares the transition to the evolution of digital payments in India, where visible interfaces mattered far less than dependable rails underneath. “The brands that win,” he says, “will be the ones that are most machine-readable—not just the most visible.”

New Ways of Advertising

As AI agents mediate shopping journeys, advertising itself begins to change. Instead of typing static search queries, consumers express intent in natural language. They describe problems, constraints, preferences, and context. Agents respond by curating options based on relevance, not keyword bids or search engine optimisation (SEO) tricks.

Google is also testing ads in AI Mode and has introduced Direct Offers. Under this new Google Ads pilot, advertisers can present exclusive offers to shoppers who are ready to buy directly within AI Mode.

At the same time, OpenAI is testing advertising in ChatGPT. This new feature complements the Instant Checkout capability in ChatGPT, which allows users to purchase products directly within the chat interface. However, adoption may take time, as users could initially be hesitant to trust an AI system to make purchases on their behalf.

Microsoft Advertising also recently published a report which says that as agents increasingly mediate shopping journeys, traditional SEO is no longer enough. Instead, retailers must adapt to Answer Engine Optimisation (AEO) and Generative Engine Optimisation (GEO).

AEO ensures that AI assistants and agents, such as Copilot or ChatGPT, can accurately retrieve, interpret, and present product information. GEO, meanwhile, focuses on making content discoverable and trustworthy in generative AI search environments, where responses are synthesised rather than ranked as links.

“AI-mediated shopping surfaces prioritise products that are well-described, consistently categorised, and clearly mapped to use cases. Poor data quality directly limits visibility, regardless of media spend,” Govindu says.

He explains that discovery, comparison, and decision-making are likely to move to AI platforms. In-interface checkout won’t replace websites or marketplaces outright, but it could weaken their grip over discovery and conversion. The competitive edge, therefore, is moving towards backend reliability, data integration, and ecosystem participation—not just front-end experience.

“When checkout becomes a capability inside every interface, storefronts stop being destinations. Brands will win by becoming the most trusted endpoint an agent can transact with—with clear rules, verifiable execution, and open interoperability,” he says. 

Similarly, Kapil Makhija, MD and CEO of Unicommerce, tells AIM that brand discoverability will increasingly depend on structured data, real-time inventory visibility, and contextual relevance rather than traditional performance advertising alone. However, challenges related to transparency, data privacy, and platform concentration must be addressed for wider adoption.

Challenges and Trust Issues 

Pandit says that for agentic commerce to work, trust and user control are foundational. He stresses users must feel firmly in control, with clear limits on what an agent can do, how frequently it can act, and how much it can spend. This is particularly critical in payments—even a single rupee matters, and confidence depends on knowing that money moves only with explicit intent and to the right destination.

Pandit argues that agents cannot function as black boxes. When an agent makes a payment, users should be able to see where the money went, why that decision was taken, and what alternatives were evaluated, as auditable reasoning is essential to building long-term trust.

Govidu highlights that as AI assumes a larger role in generating product narratives, creative outputs, recommendations, and operational choices, governance standards must keep pace with regulatory expectations on data use and intellectual property. “Even small mistakes, such as an overstated claim or skewed suggestion, can rapidly translate into outsized compliance or brand risk,” he says. 

He concludes that although AI can transact on behalf of consumers, hesitation will persist unless concerns around privacy, security, and accountability are addressed. Widespread delegation of decisions will depend on how effectively brands embed transparency, explainable logic, and meaningful mechanisms into AI-driven journeys.

As agentic commerce matures, the centre of gravity in online retail is beginning to shift. While websites and marketplaces will continue to exist, their role may evolve from destinations to infrastructure—feeding data, inventory, and fulfilment into agent-driven journeys.