The Rs 277 Crore Bet: What the MSME-TEAM Scheme's Catalogue Money Is Actually Buying, and Whether It's AI-Ready

What Rs 277 crore actually buys

The MSME Trade Enablement and Marketing Initiative, TEAM for short, was designed under the Ministry of MSME's World Bank-supported RAMP program and formally operationalized with a June 2024 set of Operating Guidelines. The headline number, an outlay of Rs 277.35 crore across three financial years, FY 2024-25 through FY 2026-27, is real and it is specific: the National Small Industries Corporation runs it as Implementing Agency, and the target is five lakh Udyam-registered micro and small enterprises, half of them women-owned, onboarded onto the Open Network for Digital Commerce.

What makes the scheme worth reading closely is not the top-line figure but the line-item mechanics underneath it, because those mechanics describe, with unusual precision for a government program, exactly what kind of digital object gets produced for that money. Onboarding and catalogue creation is capped at Rs 2,500 per enterprise, structured as Rs 50 per SKU up to 50 SKUs for consumer-facing sellers and Rs 125 per SKU up to 20 SKUs for business-to-business sellers. Account management support, covering order fulfillment, operations, logistics coordination, and B2B contract management, is capped separately at Rs 5,000 per enterprise. Logistics and packaging carry their own smaller per-order subsidies, and the Indian Institute of Packaging administers a packaging-design subsidy of 20 percent of cost or Rs 2,000, whichever is less.

A Seller Network Participant only gets paid for the catalogue work once the enterprise it onboarded has delivered at least two orders on the ONDC network. That single condition changes the incentive from "list the product" to "list it well enough that it sells," a design feature not common among digitization subsidies. The rest of this analysis turns to the harder question of whether that catalogue is built for the buyers who will actually be reading it two years from now.

The catalogue is infrastructure, not paperwork

ONDC does not run its own storefront. It runs on the Beckn Protocol, an open specification that defines a shared vocabulary for how a "catalog," a "provider," an "item," and a "price" are described so that any Buyer Network Participant, from a ride-hailing app to a grocery app, can query any Seller Network Participant's inventory without a bespoke integration. The scheme's Operating Guidelines call the deliverable an "ONDC Compliant digital catalogue" for a reason: the Rs 50-per-SKU payment is not for a product photo and a price tag, it is for a structured record that a machine on the other end of the network can parse.

That structure is the point of the whole exercise, and it is also the reason this scheme reads differently in 2026 than it would have read in 2022. Beckn's own schema registry now describes its design intent without much hedging: its catalog, item, and provider schemas, it says, are "authored with AI, for AI," and every definition is optimized for "machine navigability and agent consumption." Whether or not every field in a TEAM-funded catalogue actually reaches that standard is a separate question this analysis returns to later, but the target the guidelines are aiming at is unambiguous. This is not a listings project. It is a bet that a structured, machine-parseable product record is a more valuable asset than a merchant's own storefront, at the moment the rest of the commerce industry is making the same bet.

Why the timing raises the stakes

For most of ONDC's life, "machine-readable catalogue" mattered mainly to other apps on the same network. That changed within an eighteen-month window that overlaps almost exactly with the TEAM scheme's own rollout. In September 2025, Google published its Agent Payments Protocol, an open standard co-developed with more than sixty companies, including Mastercard, PayPal, American Express, and Coinbase, that lets an AI agent prove to a merchant that a purchase reflects a real user's authorized intent, using cryptographically signed "Mandates" rather than a human clicking "buy." In January 2026, Google and Shopify, joined by Etsy, Wayfair, Target, and Walmart, launched the Universal Commerce Protocol, an Apache-licensed open standard that lets an AI agent discover what a merchant sells and can do by querying a standardized endpoint, then complete the purchase inside a chat or search interface rather than a browser tab.

Neither protocol was built with ONDC or India in mind, and neither one currently plugs directly into Beckn. What they share with Beckn is the underlying premise: an agent acting for a buyer needs a catalogue it can query and trust, not a web page it has to interpret. That is also the premise behind the peer-reviewed research on generative engine optimization, which found that content carrying cited statistics, direct quotations, and clearly attributed sourcing was measurably more likely to be surfaced inside a generated answer than content without it. The mechanism differs, protocol-level catalogue queries versus language-model text synthesis, but the underlying requirement does not: structure and verifiable sourcing beat an unstructured claim, in both cases, because both are optimizing for what an automated reader can trust and cite.

Read against that backdrop, the TEAM scheme looks less like an e-commerce onboarding grant and more like the largest state-funded structured-product-data creation project running anywhere in the small-business economy at present, arriving in the same window that the infrastructure for agents to actually consume that kind of data went from research paper to production protocol backed by the largest retailers and payment networks in the world.

What the per-SKU cap can and cannot buy

A significant complication is that Rs 50 a SKU, capped at fifty SKUs, is not a large budget for the kind of catalogue entry that tends to earn trust from an automated reader. A rate built to cover data entry at scale is not obviously the same rate that buys a complete, differentiated, well-attributed product record, and the scheme's own guidelines seem to anticipate that gap.

What is proven, and what is still a bet

Three things in this story are documented and verifiable: the scheme's budget and mechanics, published in the Ministry's own Operating Guidelines; ONDC's scale, which had crossed roughly 49.79 million transactions within about fourteen months of its January 2023 launch and had 1.16 lakh retail sellers live across more than 630 cities and towns as of a December 2025 government statement to the Lok Sabha; and the existence, design, and launch partners of AP2 and UCP as open, published protocols with named corporate backers.

One thing is genuinely emerging rather than settled: whether a catalogue built to TEAM's cost structure, and increasingly produced by an AI cataloguing tool rather than a human, reaches the completeness and corroboration bar that agent-facing protocols reward. No public audit of TEAM-funded catalogue quality against agent-readiness criteria exists yet, and the generative-engine-optimization research this piece draws on studied general web content, not ONDC product records specifically, so the read-across is reasonable but not proven.

And one thing is openly contested: whether an ONDC-compliant catalogue built today confers a durable advantage once agent-commerce protocols standardize further. UCP and AP2 are both younger than twelve months old at the time of writing, neither is yet formally interoperable with Beckn, and the industry commentary asserting that ONDC sellers have a head start in the agentic-commerce era is, so far, argument rather than measurement.

How to read this

The scheme is real, well-specified, and larger in ambition than its trade-press coverage usually credits it for; a government program that pays by the SKU and gates payment on delivered orders is a more serious piece of policy design than "MSMEs get a subsidy" suggests. But a complete reading has two halves, not one. The first half is that Rs 277 crore is buying structured product data at a scale few private programs could match, at the precise moment the infrastructure to make that data valuable to AI shopping agents went live. The second half is that a subsidy capped at Rs 50 a line item, increasingly filled in by an automated tool, is not the same thing as a catalogue verified to be complete, differentiated, and citation-worthy to the systems that will decide whether a given business gets named in an answer or bought from directly.

Both halves can be true, and holding them together is the accurate way to read the number. The catalogue exists because of a well-designed government subsidy. Whether that catalogue is actually machine-ready is a question the subsidy was never built to answer, and it is not one a business should assume in either direction without checking.