The low-down on AI and KM with Alan Pelz-Sharpe at KMWorld 2025

As agentic AI systems increasingly start to automate decision making and knowledge work, organizations are rethinking KM to ensure accuracy, trust, and human relevance.

At KMWorld 2025, Alan Pelz-Sharpe, founder and principal analyst, Deep Analysis and Author, Practical Artificial Intelligence: An Enterprise AI Playbook, discussed how AI disrupts traditional KM—posing risks like misinformation and lost expertise—while offering transformative potential through dynamic knowledge synthesis during his session, “Role of KM in a World of Agentic AI.”

“Everything is AI. Is that really true? No,” Pelz-Sharpe said. “AI is a small component in a very big story. At the end of the day AI by itself is useless. It does nothing. What AI needs is data and if it’s going to do something it needs to know what it’s going to do and how it’s going to do it.”

Without good data, it doesn’t work. It’s a case of “garbage in-garbage out.” Data, information, and knowledge are not the same things, he said. Because of AI’s hype, everything is being labelled that when it’s not.

“Our job is to pick out the signal from the noise and that’s getting more difficult to do,” Pelz-Sharpe said.

The fundamental challenges of large language models (LLMs) are they are missing context of the information they are collecting, he noted.

Intelligent document processing is the ability to pull structured data from unstructured documents. The reality is that the one thing we know for sure is “most AI projects fail,” he said. Why do they fail? It’s not because of the AI tool, it’s because there are siloed data across the organization. There’s a lack of centralized control and barriers to cross-functional collaboration. There are difficulty scaling and innovating.

“When you hear about ‘AI agents,’ they’re just chat bots,” Pelz-Sharpe said.

Agentic AI is a type of artificial intelligence that can act autonomously to achieve a goal by planning, making decisions, and taking actions with minimal supervision. It can set its own subtasks, adapt to changing conditions, and use tools to complete complex, multi-step processes. The key word is autonomous, he noted.

“A true agent is autonomous. It choses and makes decisions for itself,” Pelz-Sharpe said.

The unstructured data pyramid is the steps knowledge managers can go through to get the correct and relevant information. This would be the process to build a small language model, he explained.

“This is one of the goals: to build those small pools and activate it,” Pelz-Sharpe said. “This is the process that every AI and agentic AI should go through.”

AI does not work correctly without human intervention, he said.

“The bottom line is, the world still has a lot of paper,” Pelz-Sharpe said. “Document processing is making it big at the moment.”

Good AI needs maintenance at the knowledge level. The “right answer” today may not be the right answer in 5 years’ time.

KMWorld returned to the J.W. Marriott in Washington D.C. on November 17-20, with pre-conference workshops held on November 17.

KMWorld 2025 is a part of a unique program of five co-located conferences, which also includes Enterprise Search & Discovery, Enterprise AI World, Taxonomy Boot Camp, and Text Analytics Forum.