Notes on product data and AI agents.

All articles

• 10 min read

  What AI-ready product data is, and what the gap costs

  AI systems compare typed attributes, not marketing copy. What AI-ready product data means, and what the gap costs in rejected listings, weak AI drafts and missed comparisons.

• 10 min read

  What a Digital Product Passport asks of your catalog

  A Digital Product Passport is a structured data record, not a label. It has to be machine-readable, kept current and readable by other systems. Many catalogs aren’t built that way.

• 12 min read

  How to evaluate a PIM in 2026: test it on your own catalog

  Feature checklists and demo screens don’t show how a PIM behaves on your catalog. Six areas to test, eight scenarios to run on your data, and the costs a quote leaves out.

• 10 min read

  ACP and UCP: the buyer’s agent reads your data, not your page

  ACP and UCP let AI agents check out on a buyer’s behalf, and agents compare products on structured data. What a buyer’s agent reads, and what your catalog has to hold.

• 8 min read

  Your supplier sends you an Excel with 47 tabs

  Supplier data cleanup is real work that nobody prices. What the hours cost, why better templates don’t fix it, and what a pipeline can take over.

• 7 min read

  Generating 10,000 AI product descriptions is the easy part

  Generating product copy with AI is easy. The work starts after: checking 10,000 descriptions for accuracy, holding one voice, and keeping every channel on the same version.

• 9 min read

  PIM vs. MDM vs. DAM: three systems, three different jobs

  PIM, MDM, and DAM aren’t competing tools. Each solves a different problem at a different layer. How to tell which ones you need, and what AI agents need from the product record.

• 8 min read

  Where the PIM sits in a composable commerce stack

  Composable commerce splits the stack. Product values still need one place that fills them in, checks them and hands each channel its own version. Where the PIM sits.

• 8 min read

  The stock is in the warehouse. The listing still isn’t live.

  Between “in the warehouse” and “on sale everywhere” sit missing attributes, hand-written copy, approval queues and channel rejections. Where each step stalls, and why.

• 7 min read

  Your pages still matter. Product data is now part of how you get found.

  Organic search still matters. More discovery now runs through feeds, comparisons and AI answers, and those read the product record: typed, consistent and current.

• 9 min read

  Running many brands on one platform without mixing their data

  Two setups are common: a PIM instance per brand, or every brand in one shared system. One multiplies infrastructure, the other mixes brand data. What a third option has to do.

• 11 min read

  One record per product, and one place that can change it

  One database doesn’t stop copies drifting. What stops it: one place that accepts writes, checks before a push, and a version of the record shaped for each channel.

• 9 min read

  Product data is what your AI systems read

  AI search, shopping agents and enrichment runs all read your product records. What an agent actually reads, and what the record needs: types, checks and a history.

• 11 min read

  Catalog completeness isn’t one score. It’s one answer per channel.

  A completeness percentage counts filled fields. It doesn’t say whether a marketplace will accept the listing, or whether AI and shopping agents have enough to work from.