2000x640 Blogartikel Headerbild Agentic Commerce V2

09. July 2026

The third shelf: your next big sales channel doesn't have eyes

Q: What is the third shelf?

The third shelf is the next sales channel after the physical shelf and the digital shelf. On this shelf, the shopper is not a person browsing products, but an AI agent buying on a person’s behalf. That changes what makes a product visible. Packaging, images and product copy matter less than structured product data: GTINs, ingredients, allergens, prices, packaging materials, certifications and sustainability attributes. In the third shelf, your product data is your packaging. If the machine cannot read it, it may never consider the product.

Q: What is agentic commerce?

Agentic commerce is commerce where AI agents help consumers find, compare and buy products. Instead of clicking through search results and product pages, the consumer gives the agent a task, and the agent evaluates the options. For brands, this creates a new decision layer between the shopper and the product. The agent does not guess. It checks whether the available product data answers the request. That means visibility in agentic commerce depends on whether your product information is complete, consistent and machine-readable.

AI agents are changing how products are found and evaluated in digital commerce. What does that mean for product data?

For fifty years, winning in consumer goods meant winning the shelf. Then e-commerce arrived and we all learned to fight for a second one, the digital shelf, with its product pages, search rankings and review scores.

Now a third shelf is opening up. And here's the uncomfortable part: the customer standing in front of it isn't a person. It's an AI agent, shopping on a person's behalf. It doesn't see your beautiful packaging. It doesn't read your lovingly crafted product page. It reads your data. All of it, instantly, and it takes every attribute literally.

At Systrion, we've spent over 25 years helping brands keep their product data straight. We've never seen a shift that raises the stakes on that boring, unglamorous discipline quite like this one.

The numbers say this is a channel, not a trend

Let's be honest: our industry has cried wolf about "the next big channel" before. So look at the data rather than the hype. Last Black Friday, AI-driven traffic to US retail sites grew 805% year over year, according to Adobe. Over the same Cyber Week, Salesforce counted $67 billion in global sales influenced by AI and agents.

And this is the early phase. Morgan Stanley projects that AI shopping agents will grow from near zero today to 126 million users by 2030, driving $190 to $385 billion in US e-commerce spending. Notably for those of us in FMCG: in Morgan Stanley's survey, groceries and consumer packaged goods top the list of what people already buy through AI. It makes sense when you think about it. Nobody wants to hand-pick dish soap. The weekly staples are the first shopping job consumers will happily give away.

The rails are being laid as we speak. OpenAI and Stripe shipped the Agentic Commerce Protocol so ChatGPT can complete purchases; Google introduced its own commerce protocol at NRF. When the payment and checkout infrastructure players move this fast, the channel is real.

What an AI agent actually "sees" when it shops

Here's where it gets interesting for anyone who owns a product portfolio.
A human shopper copes with bad data without even noticing it. If your product page has a missing attribute or a fuzzy description, she'll squint at the photo, make an assumption, maybe add it to the cart anyway. Twenty-five years of e-commerce has trained us all to shop around gaps.

An agent doesn't squint. When it evaluates products against a request like "find me a lactose-free protein pudding under two euros with recyclable packaging," it queries structured attributes: GTIN, ingredients, allergens, nutritional values, price, packaging material, certifications. If an attribute is missing, inconsistent or only exists as prose written for humans, the agent doesn't guess. It moves on to the competitor whose data answers the question.

Missing attribute, missing product. Not "ranked lower."
Not considered at all.

GS1 made exactly this point in its recent whitepaper on trusted identification in an AI-driven world: AI systems perform best when grounded in structured, verifiable and interoperable data, and without common identifiers and shared vocabularies, AI "risks amplifying noise or misinforming critical decisions instead of unlocking knowledge." Google's engineering director for commerce puts it even more directly in that same paper: agentic commerce "will require increased trust in every step that is taken on the consumer's behalf. Standards like those from GS1 are critical to identify products and ground the models."

We find a certain poetry in this. The barcode was invented so a machine could read a product at checkout. Fifty years later, the machine does the entire shopping trip. The GS1 system, with the GTIN as the anchor and GDSN as the syndication backbone, turns out to be precisely the "product truth" layer that AI agents need to ground their recommendations in something verifiable. This isn't a legacy standard scrambling to stay relevant. It's infrastructure that was built for this moment before anyone knew the moment was coming.

Meanwhile, the regulators want the same data

If agentic commerce were the only thing pulling on your product data, you could plan a comfortable multi-year roadmap. It isn't. Look at what lands in the next 18 months, just in the EU:

PPWR, August 12, 2026. Every packaging type placed on the EU market needs a signed Declaration of Conformity backed by technical documentation, and eco-modulated EPR fees start rewarding recyclable packaging. Your packaging's recyclability grade stops being a sustainability slide and becomes a number in your P&L.
ECGT, September 27, 2026. Generic claims like "eco-friendly," "natural" or offset-based "climate neutral" become prohibited without verifiable evidence. The evidence lives, or should live, in your product data.
Digital Product Passport. The EU registry goes live in July 2026, with product groups phased in through 2030. Food is exempt, but packaging data is very much in scope.
And running underneath all of it: GS1 Sunrise 2027. By the end of 2027, retail POS systems across 48 countries representing 88% of global GDP are expected to scan 2D barcodes. A QR code on a pack is a promise: scan me and get the truth about this product. But the code is only as good as the data behind it.

Notice the pattern? The AI agent, the regulator and the 2D barcode are all asking for the same thing: granular, verifiable, machine-readable data for every single SKU. Compliance data and commercial data used to live in different departments. They are now, quite literally, the same data.

The expensive words: "we'll just integrate them later"

So why aren't most brands ready?
In our experience it's rarely a lack of awareness and almost always fragmentation. The typical setup: a PIM from one vendor, a DAM from another, a GDSN data pool from a third, sustainability data in scattered PDFs and supplier emails, all stitched together with integration tape. Every hand-off between those systems is a chance for drift, delay and contradiction. Marketing's weight doesn't match logistics' weight. The recyclability grade exists in a spreadsheet but never made it into the syndicated record. Nobody notices, because a human shopper never asked.

An AI agent asks. A regulator asks. A 2D barcode scan asks. In 2026, every inconsistency that used to hide in the seams between your systems gets surfaced, at scale, by machines that don't give you the benefit of the doubt.

That's why we made what felt like a heretical decision years ago: PIM, DAM and GDSN shouldn't be three systems. Systrion runs one of only around 44 GDSN data pools worldwide, and one of the very few with a natively integrated PIM and DAM. Not because integration is a nice feature, but because we believe product data should have one home from creation to syndication. One source of product truth, feeding the physical shelf, the digital shelf and now the third one, the agentic shelf.

Where to start (hint: not with an AI project)

If you take one thing from this article, let it be this: there is no AI strategy without a data quality strategy. Companies are spending millions on AI initiatives and pennies on the data those initiatives run on. That's like buying a Formula 1 car and filling it with cooking oil.

The practical starting point is refreshingly unglamorous. Audit your GTIN-level attributes for completeness and consistency. Get your packaging and sustainability data out of PDFs and into structured, queryable form, because August 2026 is not going to wait. Make sure what you syndicate through GDSN is as rich as what you show on your own product pages. And ask hard questions about how many hand-offs your data has to survive between creation and the point where a machine reads it.

The brands that win the third shelf won't be the ones with the cleverest AI strategy. They'll be the ones whose product data tells the truth, in a format machines can trust, everywhere at once.

The machines are already shopping. The only question is whether they can see you.

Systrion AG operates a Product Information Hub with natively integrated PIM + DAM + GS1 GDSN data pool from Hamburg, Germany, and has been building the product data backbone for FMCG brands and retailers for over 25 years.

Sources:

Adobe · Salesforce · Morgan Stanley · GS1: The core relevance of trusted identification and data in an AI-driven world · GS1 US: Sunrise 2027 · Packaging Europe: PPWR-Konformitätserklärung · Cooley: ECGT-Compliance

Author:

Wolfram Koller

Founder and CEO

FAQ

What is the third shelf?

The third shelf is the next sales channel after the physical shelf and the digital shelf. On this shelf, the shopper is not a person browsing products, but an AI agent buying on a person’s behalf.

That changes what makes a product visible. Packaging, images and product copy matter less than structured product data: GTINs, ingredients, allergens, prices, packaging materials, certifications and sustainability attributes.

In the third shelf, your product data is your packaging. If the machine cannot read it, it may never consider the product.

What is agentic commerce?

Agentic commerce is commerce where AI agents help consumers find, compare and buy products. Instead of clicking through search results and product pages, the consumer gives the agent a task, and the agent evaluates the options.

For brands, this creates a new decision layer between the shopper and the product. The agent does not guess. It checks whether the available product data answers the request.

That means visibility in agentic commerce depends on whether your product information is complete, consistent and machine-readable.

Why do AI shopping agents need structured product data?

AI shopping agents need structured product data because they compare attributes, not impressions. A human can look at a packshot, make an assumption and move on. An AI agent needs data it can query.

If data is missing, contradictory, or hidden in plain text, PDFs, or Excel spreadsheets, it may not be usable at the moment of decision. In Agentic Commerce, a missing attribute can mean a missing product; the agent may skip the product entirely.

Why is GDSN important for AI commerce?

GDSN is important for AI commerce because AI agents need trusted, standardized product information. The GS1 Global Data Synchronisation Network helps brands and retailers exchange product data in a consistent and verifiable way.

With GTIN as the anchor and GDSN as the syndication backbone, brands can make their product information more reliable across retailers, channels and, increasingly, AI-driven shopping environments.

How should companies prepare for AI shopping agents?

Companies should start with product data quality, not with an AI project. The first step is to audit GTIN-level attributes for completeness, consistency and accuracy.

Ingredients, allergens, nutritional values, packaging materials, sustainability data, certifications and product claims should be structured, validated and ready to syndicate.

The real question is not whether AI agents are coming. They are.

The question is whether your product data is ready to be read, trusted and recommended by machines.

What do PPWR, ECGT and Digital Product Passports have to do with product data and agentic commerce?

PPWR, ECGT and Digital Product Passports all point in the same direction: product data must become more detailed, verifiable and machine-readable.

PPWR increases the need for structured packaging and recyclability data. ECGT raises the bar for environmental claims and the evidence behind them. Digital Product Passports create a broader expectation that product information should be accessible and trustworthy.

Compliance data and commercial data are no longer separate worlds. The same product truth that helps satisfy regulators also helps AI agents understand and recommend your products.