Every week brings a new AI headline. One day, it’s a breakthrough in AI shopping assistants, and the next day, it's a debate about whether the entire industry is sitting inside an AI bubble. But beyond the headlines, something quieter and more consequential is happening across commerce platforms.
- Product catalogs are being restructured.
- Feed systems are evolving into optimization engines.
- AI is making brands rethink how product information is created, governed, and activated.
To understand what’s actually changing beneath the headlines, we asked leaders across the commerce ecosystem, from product data platforms to infrastructure providers, what they’re seeing from the front lines. Their perspectives reveal five signals shaping the next phase of AI-driven commerce.
1. AI is exposing weak product data
For years, product catalogs were primarily built for humans (merchandisers, marketers, and content teams). AI systems, however, evaluate product data very differently.
Instead of reading descriptions the way people do, they rely on structured attributes, consistent variants, and trustworthy signals to compare and recommend products. And that shift is revealing data issues that many companies didn’t realize they had.
“AI is brutally exposing bad product data. If your catalog is messy, incomplete, or inconsistent, AI systems either recommend the wrong product or avoid recommending you altogether. At enterprise scale, fixing these issues manually simply isn’t realistic. A single broken variant or missing attribute can disrupt the entire buying journey.
As AI shopping experiences grow, the brands that succeed will be the ones with product data that machines can trust: structured, consistent, and reliable enough to support automated decision-making.”
Morten Næss,
Chief Product and Innovation Officer, Bluestone PIM
For commerce teams, this means product data quality is not just an operational concern; it’s becoming a visibility factor in AI-driven discovery.
2. Feed management is becoming a continuous optimization loop
Historically, feed management was largely mechanical. Teams prepared product exports, formatted them for each channel, and updated them periodically.
AI-driven commerce is changing that model.
With AI systems increasingly shaping product discovery and recommendations across marketplaces, ad platforms, and search feeds, feeds are moving beyond static exports and becoming dynamic systems that adapt to channel algorithms, performance signals, and shifting shopper behavior.
“In an AI-driven environment, feed management is not just about exporting the right fields in the right format. You’re orchestrating a living product graph that adapts continuously to channel algorithms and customer signals.
Feeds are becoming closer to an API than a file: refreshed frequently, personalized by context, and shaped by real-time performance.
The winning setup looks like a loop: data → activation → measurement → automated adjustments, rather than a one-way handoff.”
Reinhard Mittermaier,
Pre-Sales Consultant, Pimcore
Once feeds operate as optimization loops, performance measurement evolves as well.
Pimcore notes that as targeting becomes increasingly automated, the controllable levers shift to product data and creative quality. Instead of optimizing bids alone, teams need to measure how updates to titles, attributes, images, or variants influence performance. That shift is also redefining which metrics matter. Commerce teams are focusing more on signals such as:
- query coverage
- eligible impressions
- feed health by category
- variant-level ROAS
trbo also notes that performance measurement is expanding beyond short-term clicks. As AI influences discovery earlier in the journey, teams are evaluating how product data improvements affect discovery quality and long-term customer value.
3. The real gap isn’t missing attributes, it’s missing buying context
Many commerce teams focus on adding more attributes to their catalogs: more fields, more specs, more product details. But the bigger challenge AI systems face isn’t the quantity of data.
It’s context.
AI systems don’t just evaluate what a product is. They evaluate why it’s relevant to a shopper in a specific moment.
“The biggest gap today lies between static product attributes and buying context.
Your product data may describe what a product is with attributes such as size, material, specifications, but AI-driven recommendations depend on understanding why that product matters to a specific shopper at a specific moment.
To make that possible, your feeds need to become living knowledge bases that integrate user signals and behavioral context. Instead of static product lists, you create dynamic product experiences that adapt to each customer.”
Felix Schirl,
CEO, trbo
trbo also points to a scaling challenge many commerce teams are beginning to face. AI-driven personalization now spans multiple environments such as onsite experiences, marketplaces, advertising platforms, and emerging AI assistants. When these systems rely on fragmented data or inconsistent product signals, recommendations, and customer journeys quickly break down.
For commerce teams, the next priority is to connect product data with behavioral signals and real user context across the discovery journey.
4. AI agents won’t just answer searches; they will create demand
Ecommerce discovery followed a simple assumption: the shopper already knows what they’re looking for.
Search for a product → Compare results → Click through pages.
Agentic commerce is beginning to challenge that model.
Instead of searching for individual products, shoppers can increasingly describe a goal or situation, and AI systems assemble recommendations around it.
“Agentic commerce is enabling entirely new shopping behaviors through what I’ve learned to call goal-based shopping. Instead of searching for individual products, you describe a goal like:
- planning a 10-year-old kid’s birthday party
- (or) getting ready for a weekend camping trip in Berlin
and the AI builds a complete basket for you. Goal-based shopping creates demand. The agent surfaces products you didn’t know you needed because you were thinking about a life event, not a product category.”
Marcel Hollerbach,CIO & Co-Founder, Productsup
When shoppers describe a goal, AI systems interpret intent, context, and constraints and assemble solutions that may include products the shopper never explicitly searched for. In other words, commerce teams may need to think less in keywords and more in themes, scenarios, and customer goals.
So what happens next?
As these signals suggest, AI isn’t just adding another commerce channel; it’s changing how products are discovered and recommended.
Well, whatever happens next, one thing is clear: the infrastructure of commerce is being rewritten - quietly, but quickly.
If you’re curious how your product data and feeds can stay ready for what’s next, get in touch with the Productsup team to explore how you can prepare for AI-driven discovery and agentic commerce.
Interested in learning more? Book a demo with our partners:
Bluestone PIM | Pimcore | Trbo
About the author
