How Are Enterprise Teams Preparing AEM Content for RAG and Generative AI Workloads?

As generative AI and Retrieval-Augmented Generation (RAG) continue gaining traction across enterprise platforms, I’ve been thinking about how current AEM content architectures will need to evolve to support future AI use cases.

Many AEM implementations were originally designed around page rendering, content reuse, and headless delivery. However, AI-driven experiences introduce additional requirements around content retrieval, context quality, metadata strategy, and knowledge organization.

Some areas we’re currently evaluating include:

• Content Fragment design and granularity
• Metadata quality and governance
• Taxonomy structures for improved retrieval accuracy
• Structured vs unstructured content strategies
• Content duplication and source-of-truth management
• GraphQL delivery patterns for AI-powered applications
• Preparing enterprise content repositories for future RAG architectures

One challenge I see is that AI systems are only as effective as the content they’re able to retrieve and understand. In many enterprise environments, years of content growth have created inconsistencies in metadata, taxonomy, and content structure that may not have been problematic for traditional websites but could significantly impact AI-powered experiences.

I’m curious how other AEM teams are preparing their content architecture today to support future generative AI initiatives.

Are organizations making changes now, or are most waiting until specific AI use cases are defined before evolving their content models and governance strategies?