Commerce Is Heading Into Another Barcode Moment
How a retail lesson from Walmart, Kmart, and UPC codes helps explain the coming fight over agentic commerce infrastructure.
Commerce is heading into another barcode moment.
The Walmart lesson everyone should remember
One of the most useful stories in Sangeet Paul Choudary’s Reshuffle is not about AI. It is about the barcode.
Choudary’s line is wonderfully sharp: Kmart treated the barcode as a shovel; Walmart used it as a treasure map. That is the whole lesson. Both retailers had access to the same broad technological shift. But Kmart mostly used the barcode to improve a local task—faster checkout. Walmart saw something much bigger: a new information architecture for retail. Choudary later made the difference even plainer, arguing that Kmart was still sourcing and negotiating store by store, while Walmart used the new data to centralize sourcing, redesign replenishment, and change the power dynamics with suppliers. If you want the short version of the whole argument, that is it: same technology, very different theory of the future. (Choudary’s framing is the cleanest version I have seen.)
The historical record supports the point. Kmart had back-end scanners in use by 1981 and required barcodes on apparel by 1983. Walmart invested heavily in UPC-based scanning through the 1980s and then did the far more important thing: it built systems around the data. GS1, the standards body behind barcode infrastructure, calls the GTIN/UPC “the most important supply chain standard in history,” and notes that barcodes are scanned billions of times a day around the world. (GS1 has the scale of that shift in one place.)
I saw a version of this firsthand when I was at Procter & Gamble.
Walmart created a mandatory connection between supplier systems and store-level demand. When a specific Pampers SKU started to run low, replenishment was not a local manager’s guess, a weekly hunch, or a phone call. It was a system event. That mattered enormously. Walmart could aggregate demand across the entire chain and negotiate better pricing. P&G benefited because steadier, smoother volume is gold in packaged goods, where margins are thin and distribution efficiency matters deeply. Parents benefited because the diapers were cheaper, more consistently in stock, and available across a fuller range of SKUs.
That is what people miss about coordination power. It does not just improve the back end. It improves the shelf. Lower prices. Better in-stock rates. Better assortment. More reliability. And once those advantages reinforce one another, they become very hard to unwind. And here’s the unlock: Kmart left intelligence in the store; Walmart moved intelligence into the system.
Commerce is heading into another barcode moment.
Table Of Contents
- The new barcode: the Product Truth Standard
- From AI-assisted shopping to delegated shopping
- The race to become the control plane
- The commerce company you probably don’t know yet—but should
- Why Rithum matters
- What Rithum would have to do to win
- The Copernican Shift
The new barcode: the Product Truth Standard
The modern equivalent of UPC is the Product Truth Standard, or PTS: structured, reliable, machine-readable product truth.
The term is mine. The need is not.
OpenAI’s commerce documentation asks merchants to provide structured product feeds so ChatGPT can index products, understand core attributes, and present accurate price and availability. Google’s Universal Commerce Protocol is built around a similar premise, creating a common language for discovery, buying, and post-purchase support across agents and systems. Stripe’s Agentic Commerce Suite is pushing in the same direction from the payments layer, with near real-time product, price, and availability data delivered to agents through a single integration. And Morgan Stanley makes the practical requirement especially clear: for agentic commerce to work reliably, merchants need accurate information on inventory, pricing, variants, delivery windows, and return policies in formats machines can query programmatically.
To count as PTS, a product has to declare at least five things in machine-readable form:
- Identity: what the thing actually is.
- Attributes: what helps a machine match it to intent.
- Availability: price, stock, variant, and delivery reality.
- Policy: returns, merchant-of-record logic, warranty, subscription rules, and constraints.
- Actionability: what lets an agent actually cart it, pay for it, track it, and reorder it.
If UPC told a scanner what an item was, PTS tells an agent what it is, whether it is available, whether it can be trusted, and whether it can be bought right now.
That is a much bigger leap than most people realize.
From AI-assisted shopping to delegated shopping
We are still in a transitional phase. Many people are using ChatGPT, Gemini, and other AI interfaces the way they once used Google: to research, compare, interpret, and narrow choices. And that alone is already shifting behavior. IBM and NRF found in early 2026 that 45% of surveyed consumers already use AI during their buying journeys; 41% use it to research products, 33% to interpret reviews, and 31% to hunt for deals. Adobe reported that traffic from generative AI tools to retail sites jumped 693.4% during the 2025 holiday season. And Rithum’s own March 2026 research found that 70% of respondents had used an AI tool for shopping-related activity in the prior three months, even though fewer than 15% had completed a purchase directly through an AI tool.
But the end state is not merely better AI-assisted search. The end state is delegation.
McKinsey estimates that AI agents could mediate $3 trillion to $5 trillion of global consumer commerce by 2030. Morgan Stanley estimates that agentic shoppers could account for $190 billion to $385 billion in U.S. e-commerce spending by 2030, with groceries and consumer packaged goods likely to be one of the earliest and biggest unlocks. IBM and NRF also found that 30% of surveyed consumers want smart homes with AI personal shoppers and autonomous delivery. In other words, the next phase is not just “help me decide.” It is “handle this for me.”
I wrote recently in Agent Accessibility that humans increasingly do not operate software. They brief it. Commerce is one of the clearest places where that shift becomes practical. Retailers are still building most shopping experiences for human eyes, human patience, and human clicks. Agents do not experience a page that way. They parse structure, policy, permissions, declared actions, and machine-readable state. A retailer can have a beautiful site and still be almost unreadable to the systems that will increasingly decide what gets surfaced, compared, trusted, and bought.
That is why agent accessibility is not some side issue. It is a coming requirement.
The race to become the control plane
Once you see the shift, the strategic question changes.
It is no longer: who has the best ecommerce front end?
It becomes: who becomes the control plane that helps the whole system work?
Amazon is an obvious contender. Its Buy for Me feature lets customers discover select products from other brands’ sites inside Amazon’s shopping app and, in some cases, have Amazon complete the purchase on their behalf. Walmart is pushing in a similar direction, partnering with OpenAI so customers can shop Walmart directly inside ChatGPT. Shopify is moving from the merchant side, co-developing the Universal Commerce Protocol with Google so merchants can transact across AI channels. Stripe is attacking the problem from the execution layer, with its agentic commerce tooling designed to help businesses expose product, price, and availability data to agents and accept payments through a single integration. Google and OpenAI are defining large parts of the interface and protocol layer itself.
All of those players matter. But they do not occupy the same position.
Amazon and Walmart are sovereign retail ecosystems. Shopify is a merchant operating environment. Stripe is a transaction and payment rail. Google and OpenAI may shape the interface and protocol layers. There is still an open role in the middle of the stack: the neutral third-party layer that makes product truth, inventory truth, policy truth, and fulfillment truth usable across all of those environments.
That is the opening that matters most.
The commerce company you probably don’t know yet—but should
Most people do not need an introduction to Amazon, Walmart, Shopify, Stripe, Google, or OpenAI.
Rithum is different.
And that is part of why it is interesting.
Rithum says it was founded in 1997 as CommerceHub, and that the current company was formed when CommerceHub rebranded in December 2023 and brought together ChannelAdvisor, Dsco, and Cadeera under one name. Today it describes itself as “the connected commerce standard,” serving over 40,000 global brands, suppliers, and retailers and powering more than $50 billion in annual GMV. It says it supports the commerce journey from product listing and discovery to order fulfillment and performance optimization.
The scale and scope are larger than many people realize. Rithum’s platform spans marketplace listings, inventory management, order management, dropship, private marketplaces, shipping optimization, delivery date prediction, retail media, paid search, and commerce insights. Current company materials and a later Stripe partnership announcement describe a network spanning hundreds of marketplaces and retail destinations, 41,500+ brands and retailers, and global commerce operations across a very large share of the ecosystem.
In other words, this is not a point tool. It is already a broad coordination layer.
If you have been around ecommerce for a while, there is a decent chance you knew one of its former names before you knew the current one. Even now, Rithum’s login pages still surface ChannelAdvisor, Dsco, DemandStream, and OrderStream. That is often what infrastructure looks like before the market fully understands what layer it occupies.
Why Rithum matters
The case for Rithum is not that it already owns consumer demand.
It does not.
The case is that it already sits very close to the layer where agentic commerce becomes operational.
Rithum’s own current materials are unusually aligned with the coming bottleneck. The company says it helps brands and retailers deliver “LLM Feeds,” structured real-time product data directly into AI platforms; keep listings accurate and consistent across channels; sync inventory across warehouses and stores in real time; and use RithumIQ for classification, mapping, and data hygiene so products stay correct “from marketplaces to LLM agents.” Its product feed materials say merchants can add SKUs once and syndicate everywhere from a single source of truth. The Stripe partnership makes the positioning even clearer: structured product data, availability accuracy, order orchestration, and secure payments are being pulled together to make products discoverable by AI agents.
This is where the Walmart analogy gets interesting again.
Walmart’s genius was not that it had barcodes. It was that it reorganized retail coordination around what barcodes made possible. Rithum’s latent opportunity is similar in shape. It does not own one retail kingdom. It sits between many of them. Its opportunity is to aggregate, normalize, and serve product truth across a network rather than inside a single chain. If Walmart used UPC to centralize buying power inside one empire, Rithum could use PTS to centralize truth across many.
That is a different kind of leverage—and potentially a very valuable one.
But proximity is not destiny. Rithum is not yet the control plane.
To become it, it would have to make a much bigger move.
What Rithum would have to do to win
1. Make PTS the product
The first move is the most important one. Rithum has to stop looking like a company that helps merchants publish data to channels and start looking like the company that helps merchants declare truth once and serve it everywhere. Product identity, attributes, price, availability, policy, and actionability need to exist in a canonical, machine-readable form that can flow into OpenAI feeds, Google UCP flows, Stripe endpoints, marketplaces, retailer integrations, and future agents without spawning five different versions of the truth. In other words, Rithum has to move from helping merchants publish data to channels to helping merchants declare truth once for the machine-mediated economy.
2. Become protocol-native
The future is not going to run through one destination. It is going to run through a growing set of protocols and agent frameworks. Google’s UCP is already extending toward carts, loyalty linking, tracking, and returns. Stripe’s case is that businesses do not want bespoke integrations for every new agent. That is the opening for Rithum. It should not think of itself as a multichannel connector in the old sense. It should think of itself as the abstraction layer that translates merchant reality into protocol-ready commerce across every serious AI surface.
3. Make merchants agent-accessible, not just channel-ready
LLM feeds are necessary, but they are not sufficient. The control plane has to help merchants make commerce legible on the live web, not just in uploaded files. That means schema, APIs, structured product pages, machine-readable policies, and pages that agents can actually parse without heroic scraping. In other words: not just channel-ready, but agent-accessible.
4. Add trust truth to product truth
One of the strongest signals here comes from Rithum itself. In its March 2026 research, 58% of shoppers said they blame the retailer or brand when an AI recommendation contains incorrect product information, and 16% said they would avoid purchasing the product entirely after a bad recommendation. At that point, product data quality is no longer just a merchandising problem. It is a trust problem.
That is why I think the next real layer here is not just product truth. It is trust truth. A true commerce control plane will eventually have to coordinate not just what a product is, but whether the recommendation can be trusted, whether the merchant will keep the promise, and what happens after the sale. I have argued elsewhere that a Reputation OS becomes essential when machine-mediated systems increasingly shape how brands are understood before the click. The same logic applies here. The next control planes will have to govern both truth and trust.
5. Design for delegated replenishment
The long game here is not merely “Which sneakers should I buy?” It is “Never let us run out of dog food,” “Keep ketchup stocked under these rules,” or “Reorder diapers when size 5 falls below a threshold.” That is a much deeper coordination problem. It requires standing preferences, reorder thresholds, substitution logic, inventory truth, loyalty context, payment authorization, merchant-of-record logic, and post-purchase handling. That starts to look much less like channel software and much more like an operating system.
The Copernican Shift
Walmart won the barcode era not because it had access to UPCs, but because it reorganized retail coordination around machine-readable product data.
Commerce is entering a similar moment now.
The new barcode is not a chatbot or a prettier product page. It is the Product Truth Standard: structured, reliable, machine-readable product truth that agents can parse, trust, and act on.
The winner in this next era will not simply be the retailer with the biggest app or the AI company with the prettiest interface. It will be the player that makes products, policies, availability, and fulfillment executable across a fragmented ecosystem of merchants, retailers, payment rails, and agents.
That is why I think Rithum matters.
Not because it is already the control plane. It is not.
But because it is one of the few companies already standing in the right part of the stack, and doing so from a relatively neutral third-party position. If the market keeps seeing Rithum as channel-management software, it will value Rithum like channel-management software. If Rithum becomes the canonical PTS layer for machine-mediated commerce—protocol-native, agent-accessible, and trust-aware—the category changes. And when the category changes, the multiple changes with it.
From channel management to the neutral control plane of commerce.
Commerce is heading into another barcode moment.
The winners will not be the companies that bolt AI onto old retail workflows. They will be the companies that rebuild coordination around a new standard. Walmart did that with UPC. The open question now is who will do it with PTS.
UPC made products legible to systems. PTS will make products legible to agents. And the company that governs that shift will not just participate in commerce. It will coordinate it.