ai-product-strategy_skill

Overview

This skill helps founders and product teams define pragmatic AI product strategy. It focuses on choosing where to apply AI, clarifying human-AI boundaries, and planning iterative roadmaps that tolerate model uncertainty. Use it to decide build vs. buy, prioritize features, and design feedback loops that improve models over time.

How this skill works

I start by clarifying the user problem, current product context, and maturity of your AI stack. Then I surface trade-offs: build vs buy, single-model vs multi-model architectures, and where to place humans in the loop. Finally, I help you design evals, observability, and iteration plans so your product improves with usage and new model capabilities.

When to use it

• Deciding whether AI actually solves a clear user problem or is just hype
• Choosing build vs buy for core capabilities like embeddings, fine-tuning, or tool use
• Designing the human-AI boundary and failure-handling UX
• Planning a roadmap that anticipates model improvements and swaps
• Setting up evals, metrics, and data collection for model-driven flywheels

Best practices

• Start with the problem, not the AI—validate user impact before heavy engineering
• Define explicit human vs AI responsibilities and UX for failure cases
• Build modular systems that let you swap or add specialized models over time
• Instrument evals and observability from day one to measure regressions and gains
• Design for non-determinism: expect variability and plan safe defaults

Example use cases

• A SaaS product deciding whether to add an AI assistant for customer support and how to route confidence failures to humans
• A startup evaluating if they should fine-tune a model or integrate a third-party API for summarization
• A roadmap for migrating from a single LLM to a society of models (fast vs accurate vs domain-specific)
• Designing data collection and feedback loops to create a model improvement flywheel