Start with the problem, not just a prompt
A useful solution needs the reported issue, the affected target, and the evidence available about that environment. An incomplete target or unexplained symptom remains a question to resolve. AutoSolve carries diagnostic observations and available context into preparation.
Look for a relevant resolution
Prior knowledge can provide a strong starting point. A retrieved resolution still needs to fit the current environment. Reuse is appropriate when that fit is clear; adaptation can preserve the useful parts while accounting for differences.
Generate and prepare when no match fits
When no suitable solution matches, AI can generate a new response using the issue and context. The prepared work connects proposed steps to execution requirements. This is the bridge from a suggested answer to an action that can be inspected and evaluated.
Inspect the proposal and its requirements
Review the target, required access, runtime, tools, and intended outcome. AutoSolve evaluates available execution paths and surfaces missing prerequisites. A prepared solution may still be held because the target is not ready or an approval is required.
Define what success will establish
A successful process exit is not always proof that the original problem is resolved. Agree on an action-specific check: the expected observation, the scope it covers, and the result that warrants follow-up. Recovery or rollback is defined per action.
Keep the outcome useful
Capture verification, operator edits, quality feedback, and any rollback. These records can inform later retrieval and reuse. Optional curated training is a separate controlled workflow, not a promise that every run automatically retrains the model.