Experience Refinery

An Experience Refinery is a production mechanism that reconstructs explicit candidate models from documentary descriptions, formal sources, observational evidence, and other source material relevant to the declared scope, then carries qualified candidates through verification and release.

The Input

The refinery separates production inputs from the reconstruction mechanism. Targeted documentary and formal sources such as standards, manuals, regulations, textbooks, research, technical specifications, public documentation, and generalized documented cases provide traceable descriptions, constraints, and evidence about how a domain or process works. An LLM serves as a reconstruction mechanism whose learned statistical structure gives the production process a broad search and comparison capability.

Each source may contain only part of the model. One document defines entities. Another defines state transitions. An applicable mandatory standard may define binding conditions. A manual explains process order. A documented case may reveal an interaction that a general description leaves implicit, while any candidate structure derived from that case still requires comparison with broader evidence and any applicable governing sources before acceptance into a reusable model.

Refining Structure

The refinery reconstructs candidate operational structure from source material, including elements such as:

• entities and roles;
• states and state changes;
• processes and sub-processes;
• facts and required inputs;
• conditions and rules;
• actions and constraints;
• results and outputs;
• exceptions and boundaries.

LLMs are valuable in this stage because they can combine their broader learned statistical structure with targeted source analysis, compare language across many sources, and propose candidate structure quickly. Their proposals remain candidates; source support is assessed separately, and any normative authority comes from the applicable governing sources rather than from the reconstruction mechanism.

Verification

Refining does not end with generation. Candidate structure is checked against source material, internal consistency, coverage requirements, and explicit constraints. Conflicts are surfaced rather than silently blended.

The purpose is to move from plausible prose to explicit candidate structure that can be inspected and verified.

Output

The production output is an explicit model or model component with defined scope, source relationships, and verification status. It can make relevant relationships more explicit than the source corpus while remaining traceable to that corpus. Once it satisfies the verification and release requirements for its declared scope, it becomes a reusable model asset. It is more useful than a summary because it can guide repeated action.

The Experience Refinery is therefore the industrial mechanism at the center of Experience Capitalization: documentary descriptions, formal constraints, observational evidence, and other relevant sources are processed through AI-assisted reconstruction, explicit representation, verification, and release, resulting in reusable models.