Descriptions:
Dylan Davis, who runs an AI consultancy delivering live build sessions with business owners, presents a five-step methodology for identifying exactly where AI fits into any role or business function. The central thesis is that prompt quality is rarely the bottleneck — the real problem is that most organizations have never decomposed their roles into steps granular enough for AI to follow reliably.
The framework begins with listing every activity inside a target role (typically 20-30 items), then prioritizing three to five based on two criteria: quick wins (simple, repetitive tasks with clearly defined steps) and big time savers (high-frequency tasks where automating even partial steps returns hours per week). Davis recommends using a fast, lower-cost model — GPT Instant or Claude’s smaller tiers — to run a structured questioning session that surfaces implicit steps rather than writing them manually. The third step is explicit input/output mapping: specifying what file or data format comes in and what format the AI must return.
The final ranking step scores each candidate activity on data readiness, step clarity, and time savings, creating a sequenced roadmap that avoids the common mistake of starting with complex or infrequent tasks. Throughout, Davis emphasizes eliminating vague language (words like “realistic” or “appropriate”) and replacing it with defined criteria the AI can apply consistently. The methodology is aimed at non-technical managers and business owners rather than developers.
📺 Source: Dylan Davis · Published April 22, 2026
🏷️ Format: Workflow Case Study
