ai-pm-method-ai-engineering

What problem does it solve?

This Skill addresses the lack of structured, evidence-based decision frameworks for AI product managers evaluating AI feature feasibility, selecting the right technical implementation approach (prompt, RAG, fine-tuning, or agent), and defining valid release criteria, eliminating common costly pitfalls like unvalidated fine-tuning or unmonitored unstable agent deployments.

Core Features & Use Cases

• 3-Layer AI Stack Evaluation: Quickly identify whether an AI issue belongs to the application, model, or infrastructure layer, and assign ownership accordingly to avoid misaligned technical solutions.
• Technical Approach Decision Tree: Get clear, rule-based guidance to choose between prompt engineering, RAG, and fine-tuning based on specific failure modes, removing guesswork from solution design discussions.
• Agent Deployment Gatekeeping: Use a 5-question risk assessment to determine if an agent workflow is worth building, preventing over-engineering of unstable, high-cost agent systems.
• Release Readiness Checklists: Access 3 detailed, book-sourced checklists for model selection, evaluation set design, and pre-launch validation to ensure AI features meet both quality and business requirements. Use case: When your team proposes building an AI resume polishing tool, use this skill to route the evaluation to the relevant frameworks, assess if an agent is appropriate for the workflow, and validate all release criteria against the pre-launch checklist before development starts.

Quick Start

Use the ai-pm-method-ai-engineering skill to evaluate whether the proposed AI customer support chatbot should use RAG or fine-tuning for its knowledge base, and generate a draft evaluation plan to include in the PRD.