MCP and Agentic Workflows: Practical AI for Product Data

Agentic AI is not a future roadmap item for PIM teams. It is the current architectural decision that will decide which organisations stay visible in AI-mediated search and commerce. In this PIMvendors.com webinar, Stephan Spijkers (Co-Founder, PIMvendors.com) sits down with Morten Næss (Chief Product & Innovation Officer, Bluestone PIM) to move past the buzzwords and into what agentic AI and MCP mean for product data teams building systems today.

The session covers four layers of AI readiness in PIM, why product data quality is now binary in AI-driven search, a live demonstration of Bluestone PIM's MCP tooling, and a working example of a custom UI built with Claude Code in a single afternoon - a capability that did not exist before Christmas.

Speakers:

Stephan Spijkers - Co-Founder, PIMvendors.com

Morten Næss - Chief Product & Innovation Officer, Bluestone PIM

Watch the full webinar:

Key takeaways

• AI readiness in PIM operates across four layers, and most organisations only work on the first. The layers run from content enrichment and translation, through agentic workflows inside the PIM, through the API and MCP layer that surfaces product data to external agents, and finally to the IT architecture that decides whether any of that scales. Organisations investing heavily in layer one - better descriptions, faster translations - while ignoring layers three and four are optimising for a workflow that agents will automate within two years anyway.

• Agentic commerce makes product data quality binary. Human shoppers fill in gaps and tolerate missing attributes. Agents do not. An AI evaluating products for a buyer scores on completeness, structure and relational context - and if a product does not meet the threshold, it is not surfaced at all. There is no page two in an agentic buying journey. Organisations that have not started closing data quality gaps are already losing visibility they cannot yet measure. That is the same catalog-debt argument we make in the product data problem no one is fixing fast enough, and the practical limits of AI enrichment in what AI inside a PIM actually does.

• MCP is the protocol that connects AI agents to PIM systems - and it changes the role of the UI. Model Context Protocol gives AI agents structured, permissioned access to PIM functionality without requiring a human to log in and operate the interface. Morten's live demonstration shows Bluestone PIM's MCP tools in action: an agent querying, enriching and acting on product data directly. The practical implication is that the traditional PIM user interface becomes one access mode among many, not the primary one. PIM vendors that do not expose solid headless APIs and MCP tooling will find their systems sidelined as organisations build agentic workflows around them.

• Vibe coding has collapsed the barrier between PIM data and custom user experience. Stephan built and deployed a functioning application over a long weekend using Claude Code, with no active coding experience for the prior fifteen years. Morten's team built a custom product UI on top of Bluestone PIM's headless API in an afternoon. Neither outcome was realistic before the current generation of AI coding tools. The strategic implication for PIM buyers is direct: a composable, API-first PIM is now a prerequisite for teams that want channel-specific experiences, agent-facing interfaces or custom workflows without waiting on vendor roadmaps. When you score vendors, put API quality next to usability; how to select the right PIM covers that integration criterion in the selection method.

• SaaS PIM is becoming a toolbox, not a destination. Organisations should evaluate PIM solutions less on the quality of their out-of-the-box UI, and more on the quality of their API surface, their MCP tooling and their governance model for data accessed by agents. The UI is temporary. The data layer and its accessibility are what compound over time. CPOs and CTOs who treat this cycle as a UX upgrade are solving the wrong problem.

• The time to start is now, and the entry point is experimentation. Morten's closing advice: start talking to Claude. If you tried AI tools months ago and found them limited, try again - the ecosystem moves fast enough that the experience has changed. Stephan's framing: build the tools for future you. The foundation being laid now, through headless APIs, MCP integrations and agentic workflows, is what the next version of your organisation will operate on. CPOs, CTOs and CIOs who are not running their own experiments are building their intuition on stale assumptions.

If you want help matching an API-first shortlist to your requirements, see what PIMvendors is and how it helps.