9.3.1 — LLM-Enhanced AI Decisions

[SorraTheOrc]

Implement LLM-enhanced AI strategy layer that calls LLM service for complex choices with budget controls and fallbacks.

Acceptance Criteria:

• Hybrid strategy routes routine actions to rules and complex choices to LLM
• Budget enforcement prevents runaway costs
• Scenario tests compare rule-only vs hybrid
• Telemetry distinguishes rule vs LLM decisions
• Docs cover prompts and trade-offs

Priority: Medium

Dependencies: LLM service integration (✅ complete), rule-based strategies (✅ Task 9.2.1 complete)

Risks & Mitigations:

• Risk: LLM decisions nondeterministic or low quality. Mitigation: Constrain prompts, use structured outputs, log examples.

Next Steps:

1. Define LLM decision triggers and prompts
2. Implement hybrid strategy layer
3. Add scenario tests and telemetry

Context: Task 9.2.1 (Rule-Based AI Action Layer) is now complete with 112 passing tests. The foundation for LLM-enhanced decisions is ready.

See .pm/tracker.md for details.