Amazon Pitches Agentic Selling to Two Million Sellers
Posted on 10th Nov 2026 11:11:58 in Business, Digital Marketing
Tagged as: Amazon seller AI, agentic AI India, e-commerce sellers, SME technology, Great Indian Festival
Amazon has spent a decade persuading small Indian businesses to start selling online. Its next pitch is about what happens after they log in. In an interview with Business Standard published on November 10, 2026, Abhijit Kamra, the director overseeing seller experience across Amazon's emerging markets, said the company plans to introduce AI agents that can identify business opportunities and act on behalf of sellers. "The next shift is agentic AI," he said, while stressing that sellers will keep the final say on important business decisions.
Amazon frames the change as a move "from being a marketplace where businesses sell to becoming a long-term partner". The company says it wants to extend the benefits of AI to 15 million small businesses by 2030, and it is tying that ambition to the roughly two million sellers it already works with in India. For a shop owner deciding whether to invest another year in an Amazon catalogue, the details behind that sentence are worth reading closely.
Agents, Not Another Dashboard
The distinction matters more than the vocabulary. The seller tools in use today are mostly dashboards: they report what has already happened, in sessions, conversion rates, returns and days of inventory. An agent is supposed to close that loop, scanning demand signals, proposing the next action and, in Amazon's description, acting on the seller's behalf. Kamra's emphasis on human sign-off is a signal about how the company intends to sell the idea to merchants who are wary of handing pricing and stocking decisions to software.
There is already a base to build on. Amazon's free Selling Partner AI assistant, available round the clock in English and Hinglish inside Seller Central, was launched in June 2026 to handle onboarding, listing creation, inventory flags and export queries. It runs on Amazon Bedrock with Amazon Nova and Anthropic Claude models behind it, and Amazon says interactions with the assistant have grown 7.5 times year on year.
What the Festive Season Actually Showed
The strongest argument for putting AI in front of small sellers is not the technology but where this year's growth came from. Amazon's Great Indian Festival (AGIF) numbers, shared alongside the interview, point to sellers outside the big cities:
- More than 70 per cent of businesses participating in this year's AGIF came from tier-2 and tier-3 cities.
- More than half of Amazon's sellers are based outside metropolitan cities.
- Participation by small and medium businesses rose nearly 30 per cent year on year to a record high.
- In the first week, sales by sellers grew 3.5 times compared with business-as-usual levels, while new sellers recorded more than 2.5 times growth in units sold.
- The number of SMBs crossing Rs 10 lakh in sales grew more than 4.6 times. New sellers crossing Rs 1 lakh rose 2.4 times, and those crossing Rs 10 lakh grew more than four times.
- Shopping volumes across the festival rose more than 50 per cent year on year.
Amazon's community programmes tell the same story. Karigar and Saheli, its artisan and women-entrepreneur initiatives, say they have each reached more than 18 lakh people, and Local Shops on Amazon covers more than 3.5 lakh neighbourhood businesses. In the first week of the festival, select women entrepreneurs under Saheli recorded a 10-fold increase in sales, while Karigar sellers grew more than 1.5 times over business-as-usual levels. New partners this year include the Army Women's Welfare Association, Maharashtra's Mahila Arthik Vikas Mahamandal and the National Cooperative Organics.
Aruna Daryanani, director of Selling Partner Services at Amazon India, pointed to women entrepreneurs in Churu and Bhopal and artisans in Jodhpur and Saharanpur reaching customers beyond their local markets. That is the demand-side case for agentic tools. A seller in a tier-3 town does not have a category manager or an analyst on staff, and an agent that reads demand signals is a partial substitute for both.
The Tools Already in Seller Central
Before waiting for agentic features to arrive, it is worth knowing what is already free and running:
- AI listing tools are used by more than 200,000 sellers. Amazon says they can cut routine listing time by up to 70 per cent, filling attributes, suggesting categories and calculating fees from a product image, a URL or a short description.
- Product Opportunity Explorer is used by more than 28,000 sellers to identify demand before they buy stock.
- Sale Event Planner was used by more than 69,000 sellers to prepare for festive sales events.
For a small catalogue, that can be the difference between listing 40 products in a week and listing 200. The caveat is that an AI-drafted listing is only as good as the attributes fed into it. Titles, sizes, ingredients, warranty terms and compatibility details decide whether the listing converts or generates returns, and the assistant will happily produce a tidy listing from an incomplete input.
The Margin Story Behind the AI Pitch
Agentic tools matter less if the unit economics do not work, which is why the fee changes announced alongside this pitch deserve equal attention. Amazon says it expanded its zero referral fee offering this year to more than 12.5 crore products priced below Rs 1,000, and reduced charges under its Easy Ship service. During the first 24 hours of the festival, the number of sellers offering products below Rs 500 who recorded a 10-fold increase in sales grew more than two times year on year.
Vikram Deshpande, director of product management at Amazon India, described the fee cuts as a way to help businesses "list more selections, offer sharp and attractive prices and reinvest in their businesses". The competitive context is not subtle. Flipkart extended zero commission for fashion sellers to all price points in July 2026, having earlier covered only products priced at Rs 1,000 or less. Commission structures have become a live negotiating point between platforms courting the same seller, and a seller who does not re-check referral and shipping fees every quarter is quietly donating margin.
Agents Advise, Sellers Decide
The buyer-facing half of agentic commerce is messier, and it explains why Amazon is cautious about automation it does not control. Reuters reported in November 2026 that Amazon blocked Meta's shopping agent Muse from its store in September, saying it had accessed the site without authorisation and stored customer credentials. Amazon also won a court order against Perplexity's shopping agent Comet in March, though an appeals court overturned that ban in August. Walmart went the other way, building infrastructure in October to let third-party agents transact with shopper approval and spending limits.
The stakes are large. Morgan Stanley researchers estimated in July that shopping agents could account for 10 to 20 per cent of US e-commerce spending by 2030. For an Indian seller, the practical implication is uncomfortable but simple: if a growing share of purchases begins with a machine reading listings rather than a person scrolling a banner, then structured attributes, accurate stock counts and clean pricing data become the front door of the shop. An agent cannot be persuaded by good photography alone; it compares specifications.
What Small Sellers Should Do Now
- Clean the catalogue before automating it. Spend one week fixing attributes, dimensions, material and compatibility fields on your top 20 SKUs. Every AI feature above inherits those errors.
- Use the free assistant for the repetitive 70 per cent. Bulk listing, fee estimates and export paperwork are exactly the tasks it handles well. Do not use it for pricing decisions.
- Re-check fees every quarter. Zero referral fee slabs, Easy Ship rates and category commissions move without much notice. Build the current rates into your landed-cost sheet.
- Keep a human on price and stock. An agent that recommends a discount does not know your cash position or your supplier's minimum order quantity.
- Treat each marketplace as a separate profit and loss statement. Amazon, Flipkart and your own website have different commission, return and ad cost profiles. Tools are only useful once you can see which channel actually pays.
- Write listings for machine readers too. Complete, consistent attributes are what an agent can match against a buyer's request. Vague titles worked when only humans searched.
Amazon's agentic selling plan is, for now, a direction rather than a shipping product. But the components announced so far, an assistant that already answers in Hinglish, listing automation at 200,000 sellers and demand data at 28,000, show the company is building the plumbing. Small sellers who get their catalogue data in order now will be the ones who benefit when the agents arrive, rather than the ones whose listings the agents quietly skip.
