Foundational Concepts

Case-Based Reasoning

Brian Guy (A.A.S.) and Aurora Guy, Solutions Prodigy.

Start Here: The Short Version

This method helps a computer remember old problems and the fixes that worked. Think of each old problem as a case in a file. Each case holds the problem, the fix, and the result. A new problem arrives. The system looks for an old case that looks most like the new problem. The system changes the old fix so the fix fits the new problem. People test the fix and check the result. The system then saves the new case in the file. The next problem starts with one more case to use.

What it can do

• It finds the closest old case fast. The system compares the new problem with past cases and ranks the best matches first.
• It reuses a fix that worked before. Staff start from a proven step and adapt that step for the new facts.
• It learns from each new result. A good result adds a strong case, and a poor result adds a clear warning.

Where people can use it

• A help desk team sees the same questions each week. The system brings back the answer that solved the same question before.
• A server team sees the same faults and alerts again and again. The system brings back the repair steps that fixed the fault last time.
• Lawyers, doctors, and repair staff use past cases in daily work. A lawyer checks old rulings, a doctor checks past patients, and a repair worker checks past jobs.

Problems it can help solve

• Teams waste time when teams solve the same problem again and again. This method saves the fix so the next person finds the fix fast.
• Most search tools find text, but the tools do not learn from the result. This method saves the result, so search gets better with each use.
• New staff do not know the old cases that senior staff hold in memory. This method puts that memory in a shared file that all staff use.

The rest of this brief gives the history, the method, the 2026 gap, and a first project plan.

The Project

Goal. The desired end state is one case memory service in daily use by agents in one incident domain. The service holds solved cases with results. The service completes all four steps of the cycle on every new case.

Proof target. The project proves one claim against a retrieval-only baseline. Agents with the full cycle resolve repeat incidents with fewer steps and fewer wrong fixes. Results improve on the second occurrence of a problem because the service retains the first result.

Product at the end. The product is the Case Memory Service, version 1. The service contains a structured case store, a similarity search, an adaptation step, a result check, and a retention step with success and failure labels. Agent frameworks call the service through one API.