ai-product-strategy

An advisory skill that helps users make strategic decisions about AI products, drawing on frameworks attributed to 94 product leaders and AI practitioners. It is meant for situations such as building an AI product, deciding where to apply AI within a product, planning an AI roadmap, evaluating build versus buy for AI capabilities, or figuring out how to integrate AI into an existing product. Rather than prescribing a single playbook, it offers a structured way to reason about problem framing, architecture, and iteration.

The approach has four steps: understand the context by asking what the user is building and where they are in their AI journey, clarify the problem to distinguish genuine user needs from "AI for AI's sake", guide architecture decisions across build versus buy, model selection, and human-AI boundaries, and plan for iteration by emphasizing feedback loops, evals, and building for rapid model improvements. These are reinforced by core principles quoted from named practitioners, including starting with the problem rather than the AI, defining the human-AI boundary as a core PM decision, building for the slope rather than the snapshot, designing for "squishiness" because even high accuracy can still punch the user in the face occasionally, favoring flywheels over first-mover advantage, treating products as a society of specialized models, and using deterministic tools where they outperform LLMs.

To put the principles into practice, the skill supplies guiding questions such as what specific user problem AI is solving, what the AI should decide versus humans, how the 5% failure cases are handled, what feedback loops will improve the system, and whether evals and observability are in place. It also calls out common mistakes to flag, including AI for AI's sake, single-model thinking, ignoring failures in the UX, static architectures that cannot evolve, skipping evals from day one, and over-automation that removes humans where they add value. For deeper study it points to references/guest-insights.md containing all 179 insights from 94 guests, and lists related skills such as Building with LLMs, AI Evals, Evaluating New Technology, and Platform Strategy.