Retailers and brands rushing to deploy AI agents risk creating a new generation of disconnected technology, unless they address the product data, governance, and operating models required to make multiple agents work together, according to a new whitepaper from Akeneo.
The Agentic PIM Playbook examines the transitional phase as organisations move beyond individual AI experiments and begin deploying multiple specialised agents across the product lifecycle.
Akeneo argues that while the first AI agent may solve a relatively straightforward use case, the challenge changes fundamentally as organisations deploy more of them. As the whitepaper states, “Your first agent is a use case. Your tenth is an operating model.”
The paper identifies ten potential agents spanning the product lifecycle, including product modelling, supplier data, enrichment, content creation, localisation, asset intelligence, quality and compliance, activation, market intelligence, and insights.
While individual AI agents are likely to become widely available, Akeneo argues this accessibility to the technology itself will therefore provide relatively little competitive differentiation. What will matter is whether agents can share trusted data, business rules, workflows, and governance.
The warning comes as retailers and brands increasingly explore agentic AI to automate activities that have traditionally required significant manual intervention. Rather than relying on a single general-purpose AI assistant, Akeneo expects product teams will increasingly work alongside fleets of specialised agents, each responsible for different elements of managing, enriching and distributing product information.
But adding more agents also increases the consequences of poor underlying and unstructured data. “AI does not fix fragmented product data. It scales what it is given,” the whitepaper warns.
Akeneo identifies five foundations required to support this new environment: trusted product data; business context and repeatable intelligence; human expertise in the loop; channel intelligence; and trust, traceability and authority.
As the number of agents grows, situating these foundations correctly becomes increasingly important, and requires four shared capabilities. Agents need to work from the same governed product record; tasks and approvals need to occur in the correct sequence; taxonomies, policies and channel requirements need to be governed by shared rules; and businesses need a clear record into what has changed, why, and who, or what, made these changes.
This becomes particularly important as AI changes not only how businesses manage products, but also how consumers discover them.
As search and shopping become increasingly AI-mediated, products will need structured and authoritative information that can be understood not just by people visiting ecommerce sites, but by AI systems interpreting consumer intent, comparing products and potentially making recommendations or purchasing decisions.
The result, Akeneo argues, is that product information management is moving beyond the question of how businesses can use AI to create content faster. The larger challenge is establishing an authoritative product information environment within which multiple AI agents can operate safely and effectively.
Human expertise also remains central. Rather than removing people from product information management, the whitepaper argues that agentic AI changes where their value lies. AI can undertake repetitive work at scale, while people remain responsible for business context, exceptions, judgement and accountable decisions.
For retailers and brands, this means looking beyond the next process to automate. They need to consider how today’s individual AI projects will eventually operate as part of a much larger product intelligence ecosystem.
The whitepaper suggests organisations start with a simple question: if the business could automate one part of the product lifecycle tomorrow, which would create the greatest impact?
Justin Thomas, VP Sales – EMEA North, Akeneo said, “The interesting question is rapidly moving from, what can an AI agent do for us? to, what happens when we have ten of them? Individual agents will become increasingly easy to deploy. The harder problem is making sure they are working from the same trusted product information, following the same rules and operating within a framework the business can understand and govern.”










