I recently spoke with Patrick Knight, the CEO of IDEA, and we had an insightful discussion about something that has been around for a long time before the advent of AI: product data.
Patrick has been at IDEA for a bit more than a year, having previously worked at Rexel, and when he joined IDEA his main concern was growth and technology. IDEA has been involved in the electrical industry for nearly three decades and has helped manufacturers and distributors to resolve one of the industry’s long-standing problems, how to make product information consistent, accurate and easier for different systems to use.
What was interesting in our discussion was that this long-standing problem has suddenly become relevant in the context of AI.
For many years it has been stated that the distributor’s digital storefront is now acting as the new branch. This was reasonable since customers were going online, the sales teams were making use of digital tools, and the distributors were putting money into e-commerce, customer portals, product search, and improved digital experiences. However, I believe we are now entering a different stage. The following point of contact between a distributor and a customer might not in fact be the distributor’s website; it could be an AI agent.
A few years ago that idea would have sounded futuristic, but the change can now be seen. In March 2026 OpenAI expanded product discovery in ChatGPT, so that people could discover and compare products via more enriched shopping experiences, and its Agentic Commerce Protocol is intended to link merchants and AI systems involved in commerce.
The website was designed with people in mind, while the following interface might be one that is built for machines.
Consider the way a contractor, a procurement manager, or a maintenance professional would look for a product these days. They might go to a distributor’s website, enter a part number, filter the results by brand, voltage, dimensions, application, or availability, open a few product page, go through the specifications, download the PDF, and then finally decide if the product meets the requirements.
We are able to understand incomplete information since we are good at filling in the gaps; for example, if one company refers to a particular figure as “operating temperature”, another calls it “temperature range” and a third includes the information in a technical document, an experienced buyer is generally able to work out what the manufacturer intends.
A different issue arises with an AI agent since it has to interpret the product in a way that is consistent enough to allow it to be compared with other products and to judge whether it meets a specific request. The agent must understand the product’s identity, its attributes, the various versions available, the price, availability, compatibility, the shipping details, and at times the technical documentation. This is the direction that OpenAI’s present commerce efforts are taking: product discovery relies on merchants providing product information in structured feeds so that products can be represented more fully and accurately within AI experiences.
That means the question distributors should be asking needs to change. Rather than just asking themselves, “Can my customer find this product on my website?” we now need to ask, “Can a machine correctly understand this product?” Those are two very different questions.
The AI Agent Does Not Care About How Impressive Your Product Page Looks
Distributors should reconsider digital commerce. For many years we have worked on making websites more user-friendly. Improvements have been made to navigation, search facilities, images, product descriptions, the checkout process, and customer portals, and all of these elements are still important. However, an AI agent does not browse a website in the way that a human does; instead it reasons about information, relationships, attributes, and context.
Let’s say a customer asks an AI agent: “I need a three-phase motor for this application. It has to satisfy these voltage requirements, work within this temperature range, and be available for delivery this week.” The agent must consider the product’s attributes, compatibility, inventory, pricing, delivery details, and possibly the manufacturer’s documentation before it can provide the customer with a useful response.
If the data is incomplete, inconsistent, outdated, or spread out among separate systems, then the agent will have less reliable information to work on. The product might be there, the inventory might be there, and the website might be there too; but from the point of view of the agent the product may be hard to understand, compare or recommend.
This Is Where Standards Become Much More Interesting
My conversation with Patrick has reinforced an idea which I have been considering for a long time. The electrical industry began working on standardizing product data well before generative AI came into the picture. IDEA was set up in response to the problem of getting manufacturers and distributors to deal with product information in a more consistent manner.
Patrick described the original problem very simply: “manufacturers were saying, ‘Hey, distributors, you’re not displaying my product data right.’ And distributors were saying, ‘Hey, you all give me crappy data,’ and it comes in thirty different formats.”
The problem is still there. Although the technology associated with it has changed, the fundamental challenge is still one that people are familiar with.
To a distributor, standardization may at times seem like a mere back-office task. Someone has to draw up a mapping of the fields, tidy up the attributes, reconcile the various naming conventions, classify the products, and ensure that the same information has the same meaning in all systems. It’s easy to view this as just plain housekeeping. AI makes this kind of work far more strategic.
When product information is used by search engines, e-commerce platforms, PIMs, ERP systems, analytics tools, recommendation engines, and increasingly by AI agents, the importance of a common product language grows. Standards like ETIM are included in this wider initiative to establish a shared method of describing products across different organisations and markets.
The aim is simple in that whenever two systems are discussing the same product attribute they should have a reliable method of understanding what that attribute means.
The Electrical Industry Has An Interesting Head Start
The electrical industry is especially interesting since, over the course of many decades, it has had to deal with the complexity of product data; it involves huge catalogues, very technical specifications, a number of manufacturers, different naming conventions, several packaging levels, cross-references, and in some cases a single wrong attribute is capable of causing problems further down the line.
The fact that history provides the industry with an interesting advantage as AI becomes more deeply embedded in commerce.
The problem is not just whether one distributor’s ERP system can understand the manufacturer’s product feed. The greater issue is whether the ERP, the PIM, the e-commerce platform, the integration layer, the analytics system, the search engine, and the AI agent can all understand the product in a consistent way.
That presents a greater data problem, but it is also a considerably greater opportunity.
That is the point at which I believe that distributors should begin to go beyond AI pilots and pose a more basic question: is our product data ready for the machines which are increasingly going to make use of it?
The next competitive advantage may be data readiness
For many years the competition in the area of distribution online was mainly about which one offered the better e-commerce experience; later on the focus shifted to integrations, customer portals, personalization, mobile experiences, and digital sales enablement.
There is now another aspect coming to light: it is becoming clear which products are easiest for machines to understand.
The importance of that question is likely to rise as AI agents are incorporated into product discovery. Shopify’s 2026 report gives an early indication of the direction in which the company is heading. It states that AI-generated traffic to Shopify stores increased eightfold from the previous year in the first quarter of 2026, and that the number of orders coming from AI-powered searches rose nearly thirteenfold. The company has also introduced infrastructure which makes use of structured catalog data so that AI systems can discover, understand, and recommend products on different AI surfaces.
Since these figures relate to consumer-commerce data, it would be wrong to suppose that the same adoption curve will hold true in the case of industrial or B2B distribution. The buying process in business-to-business situations is more complicated because it involves negotiated pricing, customer-specific catalogues, contracts, credit terms, approvals, technical requirements, and relationships all of which affect the final decision.
However, the basic direction is relevant. Discovery is approaching the conversation.
What Should Distributors Do Now?
Before an “AI commerce strategy,” there is scope for the product data already inside the organization:
- What are the major sources of product information?
- Which system is authoritative for each attribute?
- How much data is trapped in PDFs or spreadsheets?
- How consistently are names, variants, prices, and availability represented?
- How well can a machine understand the catalog without human explanation?
Those questions are valuable even if AI agents never become a major sales channel for your business. Better product data improves search, ecommerce, integrations, analytics, sales enablement, catalog management, and customer experience. It can make acquisitions easier to integrate and help different systems work from a more consistent understanding of the product.
Now there is another reason to take it seriously.
Your product data may increasingly be used to answer the customer’s question before the customer ever reaches your website.
We Are Still Early, But The Direction Is Becoming Clearer
No one knows exactly how quickly agentic commerce will change B2B distribution. But the key takeaway isn’t that AI will replace ecommerce or sales teams. It’s that product discovery is changing.
For distributors, that means getting the basics right: accurate, structured, accessible product data.
The digital shelf may be becoming invisible, but the data behind it is more important than ever.
