Details
Every organization is pursuing AI—and fast—but are they truly set up to succeed at it? Or are they at risk of their AI falling short of expectations—or even failing altogether?
Boards are pushing hard, businesses are diving in, technology teams are experimenting, pilots are in place. But what is MISSING from their AI? Most organizations have not first properly defined their end-to-end AI roadmap and put the right foundations in place, especially regarding their DATA. Most data is not ready for AI production, so how do you know if you can trust your data, and trust the output from your LLMs?
Join EDM Association as we discuss our recommended roadmap for AI optimization. We will give you a preview of the content from our new AI Optimization Masterclass which outlines how to:
- Build a modern AI data layer: including data management frameworks, data catalogue, metadata, business context, and technology architecture and tools
- Develop the AI semantic layer: leveraging knowledge graphs, ontologies, data automation and controls, and scalable data products
- Accelerate AI with graph integration: exploring enterprise-ready graph databases, LLMs and the specific upskilling needed to keep the “human in the loop”
Speakers
Post-event summary
This webinar, “AI Optimization: Building Your Data Roadmap for Trusted AI,” explored why organizations need to rethink their approach to data management as they move from experimenting with AI to deploying it at scale. Expert speakers included:
- Jim Halcomb, Chief of Research & Development, EDM Association
- Shaun Rolls, Senior Advisor, EDM Association
Jim and Shaun positioned AI as a significant shift in how organizations manage data, technology, architecture, and skills. A central theme was that AI performance ultimately depends on the quality and structure of the data supporting it. Rather than a “garbage in, garbage out” approach, the session advocated for “gold in, gold out”: applying established data management practices to create high-quality, trusted data that can support accurate AI at scale.
The discussion emphasized that successful AI adoption requires more than deploying LLMs, chatbots, or individual use cases. Organizations need an end-to-end roadmap connecting data, technology, people, skills, and AI architecture. This includes moving beyond fragmented and siloed data toward machine-readable information supported by metadata, semantic models, ontologies, knowledge graphs, and shared meaning. The speakers also stressed the importance of standardizing approaches across the organization so AI systems can consume data consistently, scale effectively, control costs, and produce more reliable outputs.
A key part of the proposed approach is an AI data landscape built around several connected layers. Organizations first need an AI-ready data foundation, including metadata management and a modern data architecture. A semantic layer then provides shared meaning through semantic models and ontologies, while a graph layer connects information and relationships through knowledge graphs and graph technologies. These foundations ultimately support the AI layer, including LLMs, chatbots, agents, prompting, and retrieval-augmented generation. The speakers argued that working only at the AI layer, without the supporting data, semantic, and graph foundations, limits explainability, integration, accuracy, scalability, and optimization.
The webinar also offered a practical approach for organizations under pressure to demonstrate AI results quickly. Rather than attempting to improve all enterprise data at once, organizations should start with a strategically important business use case, identify the data needed to support it, and work backward to establish the required architecture and technology. From there, they can develop a focused “thin slice” that validates the entire process end-to-end. As Shaun explained, “you’ve got to prove that out first, by that thin slice,” establishing that the organization has the right data, people, skills, and technology before expanding into additional use cases and demonstrating ROI.
The session concluded by introducing EDM Association’s AI Optimization Masterclass as a practical way to help organizations develop this roadmap. Rather than a traditional training course, the Masterclass is designed as an interactive workshop that takes participants through the data, semantic, knowledge graph, technology, and AI layers step by step, while allowing them to apply the approach to their own challenges. The program draws on existing EDM Association resources and frameworks, including DCAM, CDMC, and DPROD, with the broader objective of helping organizations establish data management at scale and build the foundations required for trusted, accurate AI.