Every product you have ever searched for, compared, or added to a cart arrived through a supply chain most people never think about. Not the physical one that moves boxes. The data one. The chain of spreadsheets, PDFs, supplier portals, and content management systems that turns raw manufacturer specifications into the listings you actually browse.
That chain has been held together by manual labour and duct tape for decades. Now, generative AI is rewriting it from the inside out.
The real AI revolution in commerce is not happening at the chatbot layer. It is happening in the unsexy plumbing of product data.
The Problem
A single retailer might receive product information from hundreds of suppliers, each using different formats, naming conventions, and attribute structures. One supplier sends a CSV with "colour" spelled the British way. Another sends a PDF catalogue with dimensions buried in paragraph text. A third provides an API feed missing half the required fields.
The result is a permanent state of data chaos. Merchandising teams spend enormous amounts of time normalising attributes, resolving gaps, and rewriting content for different channels. A product that appears on a website, a marketplace, a mobile app, and a print catalogue may need four different descriptions, each with different character limits and formatting rules.
This is not a niche problem. It is the foundation of modern commerce, and it has been broken for years.
What AI Can Now Do
Generative AI is collapsing what used to take teams of data analysts weeks into automated workflows that run in minutes.
According to Total Retail, AI systems can now ingest unstructured inputs, map them to a defined taxonomy, and generate structured attributes and channel-ready content in a single flow. That means a messy supplier spreadsheet goes in one end, and a normalised, enriched product record comes out the other.
The capabilities stack up quickly:
Taxonomy mapping. AI models match incoming product data against standardised category trees, resolving inconsistencies across suppliers automatically.
Attribute generation. Missing fields like material, weight, or use case are inferred from available data and filled in programmatically.
Automated copy generation. Sparse specifications are transformed into customer-ready product descriptions that align with brand voice and channel requirements.
Image recognition. Computer vision models extract attributes directly from product images, identifying colour, shape, pattern, and packaging details that were never entered manually.
Inconsistency flagging. AI catches conflicts between data sources, such as a product listed as "waterproof" in one field and "water-resistant" in another.
Product data is shifting from a back-office burden into a strategic asset.
From Back Office to Strategic Asset
Once product data is clean, centralised, and enriched, it unlocks capabilities that were previously impossible at scale. As we explored in our analysis of agentic commerce going live, the shift toward intelligent automation is already reshaping how retailers operate.
Intent-driven assortment. Instead of static category pages, merchandisers can define objectives and let AI-powered systems curate product selections dynamically. The role shifts from manual curation to strategic direction.
Scenario-based micro-bundling. AI can assemble product bundles tailored to specific inventory positions, pricing constraints, and customer segments in real time. What used to require weeks of planning now happens on the fly.
Real-time substitutions. When a product goes out of stock, enriched data allows systems to identify genuinely equivalent alternatives based on deep attribute matching, not just surface-level similarity.
Natural language interfaces. Clean, structured product data is the foundation that makes conversational commerce work. When a customer asks an AI assistant to "find a lightweight waterproof jacket under £100," the system needs rich, normalised attributes to deliver a useful answer. As we covered in our look at when the agent becomes the customer, this is where product data quality becomes a competitive moat.
The Autonomous Supply Chain Horizon
Product data is just one layer of a much larger transformation. Supply chains themselves are learning to think.
McKinsey estimates that AI could generate $190 billion in value across travel and logistics, plus $18 billion in direct supply chain operations through applications like automated documentation and fleet management systems.
The World Economic Forum frames the evolution as a three-stage progression. First, digitalisation: replacing manual processes with cloud systems for real-time visibility. Second, AI-assisted adaptability: using machine learning and simulation to anticipate disruptions. Third, complete autonomy: AI acting independently in real time.
Most organisations remain in the early stages. According to PYMNTS, IDC predicts that by 2028, half of large enterprise supply chains will achieve network-level visibility, with agentic AI cutting disruption response times by 25 percent.
The question is no longer whether supply chains will become autonomous. It is how quickly organisations can build the data foundations to get there.
What Comes Next
The companies that treat product data as strategic infrastructure will have a compounding advantage. Clean data feeds better AI models, which generate better product experiences, which drive higher conversion, which funds further investment in data quality. The flywheel is real.
The laggards will find themselves stuck in a familiar loop: hiring more people to fix the same data problems, while competitors automate their way past them.
Watch for consolidation in the product information management space as AI-native platforms challenge legacy PIM and MDM vendors. Watch for retailers demanding structured data standards from suppliers as a condition of doing business. And watch for the first wave of truly autonomous merchandising systems that do not just recommend decisions but make them.
Sources
Is your organisation still treating product data as a back-office problem, or have you started building the foundation for AI-driven commerce?