Product data management is being rebuilt around AI agents as first-class consumers of the PIM. In this episode of The Single Source, Crystallize CEO Bård Farstad joins PIMvendors co-founders Stephan Spijkers and Kees Jobse to examine the architectural shift that is already changing how product data is modeled, served, and acted on. The conversation moves from the historical role of PIM in the commerce stack to a concrete vision of what it means to have your product data available in milliseconds — structured, semantically correct, and accessible to both developers and agents through a single MCP interface.
Crystallize is a headless, GraphQL-based PIM and e-commerce platform built for developers and product teams who need real-time delivery, rich product storytelling, and flexible content modeling. This session draws directly on how Crystallize has architected its platform around Skills, AI agents, and MCP — and what that means for any organization evaluating its readiness for agentic commerce.
Speakers: Stephan Spijkers — Co-Founder, PIMvendors.com
Chris Jobse — Co-Founder, PIMvendors.com
Bård Farstad — CEO, Crystallize
Key Takeaways:
Product data must be accessible to machines, not just humans. The baseline requirement for agentic commerce is structured, semantically correct product data delivered in milliseconds. Specs, storytelling, media, pricing, and variant relationships all need to live in a single backend with a self-describing API — because if an AI agent cannot find and interpret your data inside a certain response window, it moves on. Your products become invisible before the buying journey even starts.
The integration tax of the classical PIM stack is no longer acceptable. The traditional model — PIM enriches data, pushes it to an e-commerce platform, which serves the storefront — creates a least-common-denominator problem. Data is constrained to what both systems support, flexibility is lost at every handoff, and the architecture cannot serve the channels agents actually use. A single headless backend with one API, one data model, and one MCP removes that tax entirely.
Skills and MCP are compressing weeks of work into hours. Crystallize has embedded AI at every layer of its platform — data modeling, bulk import from spreadsheets, SEO enrichment, plugin development, RFP responses, and bug resolution in Slack. Tasks that previously required days of developer time or external consulting are now completed in minutes and reviewed rather than built from scratch. The shift since late 2024 is not incremental: major features are shipping daily, not weekly.
The buying journey is splitting into two distinct tracks — and both require AI-ready data. Experiential purchases (cars, furniture, fashion) still demand brand-controlled, immersive discovery environments. Functional purchases (spare parts, consumables, commodity reorders) are already moving to agentic channels where Claude or ChatGPT filters the consideration set before the customer ever visits a storefront. Organizations need structured product data that serves both tracks simultaneously — or they are absent from one of them.
Model optimization is the lever most organizations are not yet pulling. Not every task requires the most capable model. Translation, classification, and bulk field enrichment can run efficiently on lighter models. Complex data modeling, architecture decisions, and one-shot plugin generation benefit from full model capability. Mapping tasks to the appropriate model — rather than defaulting to the most powerful option for everything — is where meaningful cost control happens at scale.
Architectural readiness is now a leadership decision, not an IT backlog item. Organizations that cannot expose structured product data through modern, agent-compatible interfaces within the next 12 months face a compounding disadvantage: their products are filtered out at the discovery stage, their internal teams cannot utilize agentic workflows, and their vendor relationships will come under pressure as customer expectations shift. The question is binary — is your current architecture ready to serve these channels today?
👉 Looking for the right PIM or composable commerce partner for your stack?
Compare vendors and book a call with our team at pimvendors.com
