![How To Scale SKU-Heavy Catalogues Without Duplication [Guide For Retailers]](https://www.bluestonepim.com/hubfs/How-To-Scale-SKU-Heavy-Catalogues-Without-Duplication-1.png)
Product Information Management
AI is now part of almost every PIM platform, but most vendors' AI stops at a suggestion: generating a description, translating a field, flagging a gap, that a person still has to review and apply. This article compares Bluestone PIM, Akeneo, Inriver, Salsify and Pimberly on what happens after that point: content generation, data quality, workflow support, Model Context Protocol (MCP), and how much of the work an AI agent can actually carry out on its own.
If you want the underlying framework for evaluating any PIM vendor's AI maturity, not just these five, see How to Compare AI Agents in PIM Systems. This article assumes that framework and applies it to named vendors directly.
What Does AI In PIM Cover Today?
AI in PIM now covers several types of product data work, but they do not provide the same level of autonomy.
| Type of AI capability | What it does | Typical example |
|---|---|---|
| Content assistance | Generates or edits product content | Writing descriptions from technical attributes |
| Data analysis | Finds gaps, anomalies or inconsistent values | Detecting missing materials or incorrect categories |
| Workflow assistance | Runs an AI task inside a defined process | Translating content during product onboarding |
| Conversational access | Turns natural-language requests into searches or actions | Finding products below a completeness threshold |
| Agentic execution | Reads the situation, selects an action and carries it out | Updating data and triggering publication after validation |
Most current PIM AI functions sit in the first three categories. They save time, but a person still starts the task, reviews the output or moves the product to the next stage.
Agentic execution changes that relationship. The agent receives a goal, works across live product data and performs the operations needed to reach it. Human control remains through permissions, approvals, workflow rules and audit records.
What Makes a PIM Agentic?
An agentic PIM lets AI take part in product data operations rather than limiting it to text generation or recommendations.
For example, an AI assistant may identify products without German descriptions and show the results. An agent can find those products, generate the missing values, write them to the correct market context, check completeness and trigger the next approved workflow step.
This depends on more than the quality of the language model. The PIM needs structured data, machine-callable operations, granular permissions and an interface through which the agent can use those operations.
A useful distinction is:
-
AI-assisted PIM: AI prepares an answer or suggestion for a user.
-
Agentic PIM: AI can act on product data within defined controls.
Why Does MCP Deserve Extra Scrutiny When Comparing These Vendors?
MCP, or Model Context Protocol, gives AI applications a standard way to connect to external systems, data and tools, so the same MCP server can serve ChatGPT, Claude, Cursor or a company's own agent without a custom integration for each.
Having an MCP server, on its own, tells you almost nothing about what a vendor actually lets an agent do. Before comparing the five vendors below, it's worth checking each one against the same questions:
- Does it support reading, writing, or both?
- Which product data concepts does it actually cover?
- Can agents start workflows or publication, or only retrieve records?
- How do authentication and permissions work?
- Is the feature live in production, in beta, or only planned?
An MCP server that only retrieves product records is useful for search and recommendations. It's a different thing entirely from an MCP layer that lets agents update attributes, manage catalogue structures, and trigger product operations directly.
How Do the Five PIM Vendors Compare?
Each vendor takes a different route, from AI-assisted content creation and data checks to connected agents that can read, update and manage product information across workflows. The main difference is not whether AI exists, but how far it can act.
| Vendor | AI functions | Model choice | MCP position | Main agentic commerce direction |
|---|---|---|---|---|
| Bluestone PIM | Autonomous onboarding, enrichment, validation, publishing and product data management | Customers can connect their preferred models | Native MCP with read and write access across core product, catalogue, asset, publishing and workflow concepts | External agents operating governed PIM workflows |
| Akeneo | Onboarding, classification, enrichment, translation and catalogue management | Supports customer-connected LLMs | MCP server in beta | Broad catalogue access, with several operational gaps |
| Inriver | Product data access, content enrichment and developer support | Inriver Inspire and embedded agents | Two MCP endpoints | Querying live data and supporting REST API development |
| Salsify | Conversational assistance and AI tasks inside PXM workflows | Customer AI providers or SalsifyIQ | No native MCP layer presented | AI works mainly within Salsify workflows and channel processes |
| Pimberly | Copy generation, image analysis and reusable AI prompts | Pimbles supports the customer's choice of LLM | No native MCP layer presented | AI supports defined content and attribute tasks inside Pimberly |
The table is based on current vendor product pages and technical documentation.
How Does Each Vendor Approach AI and Agents?
Vendor strategies differ in scope and technical maturity, ranging from AI-supported content generation and automated data validation to agent-based architectures that orchestrate product information across interconnected systems and workflows.
How Does Bluestone PIM Use AI and MCP?
Bluestone PIM combines embedded AI capabilities with a structured product data foundation, a headless, API-first architecture, and native MCP support. More than 700 task-level API endpoints provide full UI/API parity, meaning every operation available through the interface has a machine-callable equivalent.
Compatible agents can use MCP to work with core PIM concepts, products, attributes, relationships, catalogue structures, assets, market contexts, completeness, tasks and publishing, operating within the platform's existing permissions, validation rules and audit history. This runs on the same single source of product truth that validates data continuously and keeps a change history for each value, so people, connected systems and agents all work from the same version of the truth. The specific rules behind this, category assignment, workflow handoff, channel sync, are documented in the automation e-book.
Download free e-book
Automation in Product Information Management
This e-book explains how automation works in product information management, which catalogue workflows can be automated today, and how an event-based architecture creates the foundation for the agentic PIM.
Product teams remain responsible for goals, brand voice, quality standards, approval rules and the exceptions that need human judgement. The exact level of automation depends on workflow configuration, integrations and which agent tools are in use, and MCP coverage shouldn't be treated as identical to the full API surface. Bluestone PIM already provides the architectural foundation for agent-led product data operations, while end-to-end autonomous execution across every process remains the direction the platform continues to develop towards.
How Does Akeneo Use AI and MCP?
Akeneo uses AI across supplier onboarding, product classification, content enrichment, translation and data management. Its approach is focused on helping teams process and improve large volumes of product information more efficiently.
The platform provides an MCP server that supports natural-language access and read and write operations across a broad range of product data. The service remains in beta, and some areas, including media handling and workflow execution, are not yet covered. This gives agents useful access to catalogue records, but they cannot currently carry every process from data retrieval to full workflow completion.
How Does Inriver Use AI and MCP?
Inriver applies AI to product onboarding, enrichment, translation and catalogue processes. Its MCP approach focuses on giving connected AI tools access to live product data and helping developers work with the platform's API.
This makes it useful for search, recommendations, analytics, content support and validation. The scope remains narrower than a full agentic operating layer. External agents can retrieve governed product data and support development tasks, but the platform does not currently present MCP as a way to manage the entire catalogue workflow directly.
How Does Salsify Use AI for PXM and Agentic Commerce?
Salsify places AI inside its existing PXM workflows. It supports content work, data extraction, governance and user assistance as part of processes that combine automated steps with human review.
The platform's approach is centred on helping teams work faster within Salsify and distributing approved product data to external channels, including AI-driven shopping experiences. It does not currently position native MCP as the connection layer through which external agents can directly manage product data and operational workflows.
How Does Pimberly Use AI in Product Data Workflows?
Pimberly uses AI mainly to support product content, image analysis and attribute enrichment. Teams can apply repeatable AI tasks across groups of products to reduce manual work and process catalogue data at scale.
These capabilities remain part of workflows managed inside the Pimberly platform. Its agentic commerce positioning focuses on preparing structured product data for AI-driven discovery and sales channels, rather than giving external agents direct operational access to the catalogue through a native MCP layer.
Which Platform Is Most Ready for Agentic PIM?
Bluestone PIM offers the strongest architectural foundation among the five for companies that want agents to move beyond individual AI-assisted tasks into a more active role in product data operations. That position comes from how the platform is built: 700+ task-level APIs, full UI/API parity, and native MCP give agents machine-callable access to core product data and operations, with the single source of truth providing the governed control layer, lineage, permissions and audit history, that agents need to work from safely.
Akeneo comes closest in terms of MCP breadth. Its server supports many product data concepts and both read and write operations. The beta status and missing workflow execution limit how far an agent can carry a process today.
Inriver provides MCP access to live product data and tools for API development. This is useful for AI applications and technical teams, but it is a narrower model of agent participation.
Salsify brings AI into PXM workflows and sends product data to AI shopping channels. Pimberly uses AI for repeatable content and attribute tasks. Both can reduce manual work, but neither currently presents the same direct agent-to-PIM operating model.
Is Your PIM Ready for Agents to Do the Work?
AI can help teams write faster, translate more content and find data gaps. Agentic PIM changes the operating model itself.
Instead of waiting for a person to move every product through every step, agents handle supplier onboarding, enrichment, validation and publishing. People define the objectives, rules and exceptions.
Bluestone PIM is built for that shift. Agents run the daily product data work. Your team remains in control of the product truth.
