Built upon a foundation of expertise captured from subject matter experts, Xpertise Augmented Generation (XAG) directs foundational AI models toward specific problem-solving goals. XAG enables self-directed AI Constructs to fill the roles of knowledgeable actors within an enterprise.  It extends beyond the standard Generative AI prompt-response paradigm and beyond Agentic AI approaches enabling more contextually aware cognitive AI solutions such as AI Role Players, AI Instructors, AI Advisors and AI Commanders.

What are the elements of Xpertise Augmented Generation (XAG)?

Generative AI

All Generative AI models leverage prompts which are a set of tokens within a context window (e.g. Chat GPT 5 has extended its window to 400,000 tokens). The entirety of a chat conversation is leveraged as the prompt to the massive classifier (or set of classifiers) that generates the response. Many organizations have used Retrieval Augmented Generation (RAG) or Cache Augmented Generation (CAG) to leverage Generative AI. RAG takes an organization’s data sources and retrieves appropriate information to add to the prompt, while CAG attempts to add all the data to the prompt.     Neither of these approaches can leverage what is in the heads of the Subject Matter Experts.

Agentic AI

In an effort to break away from the limitations of “prompt-response” interactions with AI, many AI companies have pushed the idea of Agentic AI. This typically means that the AI is connected to a workflow within the organization and can take some actions to support that workflow, e.g., sending email or adding data to a spreadsheet. Agentic AI, however, is not “self-directed”, it does not proactively advise the user but rather assists the user through a specific workflow.   

Subject Matter Expertise

Most of the expertise found in an organization is not in the data or documents. Rather it is found in the heads of its experts. These experts have learned the hard way through trial and error and numerous failed attempts. As Niels Bohr states: “An expert is a man who has made all the mistakes which can be made, in a narrow field.” Discovery Machine has a patented methodology and technology that enables SMEs to introspect and articulate their expertise. The result is mental model representations that can lens and focus Generative AI classifiers to support conversational, goal-directed, situationally aware, and dynamically reactive AI Constructs.