Descriptions:
Dylan Davis, who runs an AI consultancy, introduces a four-question diagnostic for identifying exactly where AI breaks down in any given workflow — and why most teams that claim to use AI daily are still stuck at the lowest tier of adoption. The central distinction he draws is between “AI-assisted” (using AI on top of old ways of working, with minimal structural change) and “AI-native” (fundamentally redesigning how work is done around AI capabilities).
The four diagnostic questions are: Can AI see it? (does the AI have access to the right context in a format it can process); Can AI understand it? (has the user externalized their judgment and standards into the AI’s prompt, including binary pass/fail quality criteria); Can AI act on it? (does the AI have write access to downstream tools like CRM, email, and task trackers, or is the human still copy-pasting outputs); and Can AI improve on itself? (is the system compounding lessons and preferences over time, turning the AI from a static tool into a learning asset). Each question is illustrated with practical examples, including an email quality rubric with explicit binary standards.
Davis also argues that “AI native” scales from individual contributors up through teams, departments, and entire companies — and that most people should start by getting excellent at the individual level before trying to transform broader organizations. The framework is tool-agnostic and applicable whether someone is working with ChatGPT, Claude, or desktop agents like Codex.
📺 Source: Dylan Davis · Published May 18, 2026
🏷️ Format: Opinion Editorial
