In the eCommerce landscape, Artificial Intelligence (AI) is reshaping the game. This blog series explores how AI’s intelligent algorithms are revolutionizing online businesses, from personalized product recommendations to efficient inventory management. Join us for insights on leveraging AI to enhance the digital shopping experience and overall success in eCommerce.
ADVERTISEMENT
You have watched multiple YouTube videos on the art of running, downloaded and signed up on Strava, but wait! You still need shoes.
This means opening several shopping apps, checking product pages, comparing reviews and then making a decision.
You just witnessed a consumer journey that is turning archaic as you read this. In the new consumer journey, the runner might open an AI app and say, “Find me a pair of running shoes under Rs 10,000. Check the reviews, compare prices, make sure they can reach me by Friday and buy the best one.”
That is the idea behind the next phase of shopping, or to be exact, AI-led shopping. The technology is still at an early stage, but the change it could bring to ecommerce is easy to see. A consumer may still buy the same pair of shoes from the same marketplace. But the search, comparison and recommendation may happen with an AI agent deciding what is worth bringing back to the consumer.
For ecommerce companies, that changes where the consumer journey begins and who gets to influence the final choice.
Let’s understand this step by step. You wouldn’t want to miss the last section.
Why are marketplaces wary?
A marketplace search page is where consumers discover products, compare options and encounter sponsored listings. Brands pay to appear there because that is where the consumer is making a decision.
Subarna Mukherjee, Founder & Global CEO, Shop Culture, sees this as one of the main reasons marketplaces are putting controls around agents.
“An agent never sees the search results page, and that page is the shelf brands rent,” she said.
Amazon generated about $68.6 billion in advertising revenue in 2025, Mukherjee pointed out, while more than 70% of ad spend across her India portfolio still goes towards sponsored search.
If agents become the route through which consumers discover products, the value of that search page could come under pressure because an agent may compare products across several platforms before recommending one.
Amazon has placed restrictions on AI agents accessing its services and requires agents to identify themselves under its updated Business Solutions Agreement. Meta’s Muse has also been unable to place orders on Amazon, while other retailers and commerce platforms have chosen to work with agents.
For Chirag Taneja, Co-founder & CEO, GoKwik, the response is linked to how ecommerce has traditionally been built. “Marketplaces were built around the assumption that the consumer experience, from discovery and comparison to browse-to-buy, happens within their walls,” he said.
An AI agent can separate those stages by doing the discovery elsewhere and bringing the consumer to a platform only when it is time to transact.
That could affect impressions, sponsored listings and the behavioural data that marketplaces use to monetise their traffic.
The response from Indian platforms could therefore take several forms, from restricting access to building their own AI interfaces or working with AI companies. The last option is clearly beginning to take shape.
Swiggy allows AI-native ordering across Food, Instamart and Dineout, has partnered with Sarvam for multilingual voice commerce, and is opening its commerce stack to approved builders.
And India has one advantage that could make those partnerships easier.
India already has one important piece in place
Razorpay and NPCI have been working with OpenAI on agentic payments on ChatGPT, with BigBasket among the early commerce use cases. Claude has also been connected to commerce services including Zomato, Swiggy and Zepto through agentic payment initiatives.
This matters because an AI agent that can recommend a product but cannot complete the transaction remains an assistant. Once payment can also be handled, it starts taking on the role of a shopping agent.
Mukherjee expects partnerships between commerce platforms and AI companies to grow around this infrastructure. She points to the role of payment rails in India and the scale of UPI as reasons why platforms may find it difficult to simply wall off agents.
“In India, leverage sits with payment rail, not marketplace,” she noted.
According to her, Indian commerce players may work with AI companies around access to catalogues and commerce data rather than hand over exclusive control of their platforms.
Taneja from GoKwik expects the partnerships to take time because an agent needs much more than a payment button. Inventory has to be current, prices have to be accurate and the agent needs enough information to judge whether a seller and product can be trusted.
“AI agents need reliable structured data, real-time inventory, pricing and trust signals, along with payment rails,” he said.
Shradha Agarwal, Co-founder & Global CEO, Grapes Worldwide, sees the commercial logic for both sides. Commerce platforms have the catalogue, inventory, pricing, fulfilment and transaction history, while AI companies bring reasoning, personalisation and automation.
She expects these capabilities to come together as agentic commerce develops, much as brands had to adapt when search engines became an important route to product discovery.
The first big change may be discovery
Consumers are unlikely to hand over every purchase to an AI agent overnight. The more immediate use case is likely to be research and discovery.
A consumer looking for running shoes can ask an agent to compare products, prices, reviews and delivery timelines and return with a shortlist. Agarwal expects this part of the journey to change before agents take over a large share of transactions.
“The homepage or search bar may no longer be the starting point,” she said.
The pace will also vary by category. “The purchase types most vulnerable early on are replenishment-led ones,” Taneja said. “An agent already knows the consumer’s preferences and can reorder at the right price.”
Fashion, furniture and other high-consideration categories may take longer because consumers often want to inspect and compare products themselves.
Mukherjee also expects repeat purchases to be among the early areas of adoption. In the FMCG and pet-care businesses in her portfolio, she said, around 22% of revenue comes from repeat purchases.
Those are transactions where an agent already knows a consumer's preferences and can make a purchase based on price, availability and delivery.
There is no reliable India-wide estimate yet for how much ecommerce spending could become agent-led. Mukherjee estimates that agents could account for 5-8% of Indian ecommerce GMV by 2030.
Global estimates are higher. Morgan Stanley has estimated that agents could account for 10-20% of US ecommerce by 2030, equivalent to $190–385 billion, while Bain has put the range at 15-25%.
These are forecasts rather than established market shares. The more immediate change could happen in how consumers reach products and decide what to buy.
That also brings a direct consequence for brands. If the consumer is no longer browsing a marketplace search page, the way brands pay for visibility on that page starts to lose some of its relevance.
What happens to retail media?
Search advertising and sponsored listings work because a consumer sees products, compares them and clicks. An agent may assess a much larger set of products and return with two or three recommendations.
Taneja believes this could put pressure on sponsored search because the agent is trying to satisfy the consumer's requirements rather than simply follow the highest-paid placement. “If AI searches, it optimises user outcome, not highest bidder,” he said.
For brands, that could create a different way of thinking about visibility.
Mukherjee expects potential revenue pools around paid catalogue and API access, sponsored recommendations within AI interfaces and fees linked to completed sales instead of clicks.
“Retail media sells attention, agents have none,” she said. “They don't scroll, see banners or click a third sponsored listing.”
She has already seen clients move from ROAS (Return on Ad Spends) towards TACOS (Total Advertising Cost of Sales) and expects new measures around the cost of getting an AI agent to select a product.
That is also where GEO starts becoming more relevant to commerce. Siddhartha Vanvani, Co-Founder & CEO, DareAIsearch, says the role of GEO is moving closer to the point of purchase as AI becomes a recommendation layer.
Vanvani puts current Indian GEO programmes at around Rs 75,000 to Rs 3.5 lakh a month, with larger enterprise engagements going higher.
“Retail media will move from buying placement to influencing machine recommendations. Today, brands pay to appear higher in a marketplace search. Tomorrow, they may compete to become the product an agent selects,” Vanvani says.
Chiming in, Ambika Sharma, Product Architect at NeuroRank & Chief Strategist at Pulp Strategy, said, “Retail media will not disappear; its unit of value will change. Instead of paying mainly for a keyword position or sponsored shelf slot, brands will compete for consideration inside an AI-mediated decision.”
She added, “New revenue pools could emerge around agent-ready sponsored offers, commerce APIs, transaction fees, premium merchant services and measurement of AI-driven recommendation, assisted conversion and incrementality.”
This changes the value of being visible inside a marketplace. Brands may still need to win the search result, but they may also need to become legible enough for an AI system to evaluate and recommend them.
Getting an agent through the door is only the first step
And this is where the Indian ecommerce story gets more interesting. The industry can debate how much money will move into agent-led commerce and what happens to retail media, but all of it depends on a fairly basic requirement.
The agent needs to be able to access and understand the commerce site in the first place.
A September 2026 study by FTA Global offers a glimpse into how ready Indian ecommerce is for that reality.
FTA tested 50 Indian ecommerce sites across 10 categories using Claude, GPT and Gemini. It ran 450 tasks covering return windows, delivery charges and product prices. The agents completed 358 tasks, or about 80%.
That sounds encouraging until the failures are examined more closely.
FTA found that 21 of the 50 sites blocked or failed its generic automated crawler. Yet the real AI agents performed almost as well on many of those sites as they did on sites that were open to the crawler.
They completed 77% of tasks on sites where both a plain fetch and browser-based crawl were blocked, compared with 82% on sites that were open to the generic crawler.
The bigger problems came once the agents were on the site. Of the 92 tasks they could not complete, the most common issue was that the required answer was not clearly stated. Another 13 failures involved JavaScript content that did not load properly, while 11 were caused by a site being blocked or returning an error.
The details of product pages tell a similar story. FTA found the product name in the server-side HTML on 17 of 24 pages tested, but the price was available there on only 12. The product schema was present on 19 pages. Return policy information appeared in the schema on just three pages, while delivery information appeared on two.
Taneja believes this is where the quality of the underlying commerce infrastructure becomes important.
“The infrastructure layer is what agents will depend on,” he said. “If agentic commerce grows, the important players will be the ones whose data and checkout infrastructure agents can trust and plug into.”
The FTA findings put a number around that readiness gap. The average Agent Readiness Score across 42 fully scoreable sites was 43 out of 100. Twelve sites scored 80 or more, while 18 scored below 20. Pharmacy sites had the highest category score at 87, while grocery and quick commerce averaged 23.
In other words, the infrastructure challenge is not limited to building an AI interface or connecting a payment rail. It also sits inside the ecommerce websites that agents will have to navigate.
Where does that leave Indian ecommerce?
India is entering agentic commerce with different parts of the system moving at different speeds.
The payment infrastructure is being prepared for AI-led transactions, and AI companies and commerce players are already testing ways to connect their systems.
At the same time, FTA's research shows that many ecommerce websites still present information in ways that are designed primarily for human shoppers.
That gap will matter as agents move from helping consumers search to making more of the decisions themselves.
The marketplace itself is unlikely to disappear. It may become one part of a journey that begins elsewhere.
ADVERTISEMENT
The conversation highlighted a fundamental shift underway in the relationship between audiences, platforms and brands. Rather than treating Gen Z as a uniform consumer segment, marketers are increasingly having to account for individual behaviours, fragmented content consumption and a stronger expectation
ADVERTISEMENT
Storyboard18 Awards for Creativity 2026 honoured campaigns that blended storytelling, cultural relevance and measurable impact across platforms.
ADVERTISEMENT
Special Coverage
“Range matters now”: Microsoft’s Puneet Chandok, GrowthSchool’s Vaibhav Sisinty on the skills that will define Gen Z careers
Special Coverage
‘Somebody else will do it’ mindset must change: Tukaram Mundhe on his vision for India at Storyboard18 Global GenZ Summit
Special Coverage
‘The real pink slip in the AI era is your refusal to learn’: Microsoft's Puneet Chandok
Special Coverage
'Collar se pakad ke le gaye’: Bhuvan Bam recalls Delhi police mistaking him for a pickpocket amid fan frenzy
Brand Makers
Storyboard18 Global GenZ Summit: Shraddha Jain on authenticity and why creators should ‘design your own game’
Digital
What ChatGPT’s new ad push means for consumers, advertisers and OpenAI
Digital
ChatGPT expands ad business with visual formats, measurement and brand-suitability tools
Media & Entertainment
Can AI predict a microdrama hit? BULLET says its hit rate has risen to 20-25%
Digital
OpenAI to roll out text watermarking in EU, opens detector access to researchers
Digital
ChatGPT ads: New money for media or another fight for the same ad budget?