5 Steps to Break Roles into AI-Bite-Size Activities
Decompose roles into 20-30 activities, prioritize 3-5 quick wins or big time savers with clear steps/inputs/outputs, then build focused AI folders (Claude.md/agents.md + data) for reliable automation.
Decompose Roles into Automatable Activities
Successful AI users break weekly tasks into granular activities AI can handle, rather than relying on perfect prompts. Start by listing 20-30 activities per role—imagine strapping a GoPro to yourself and cataloging everything observed. Prioritize 3-5 using two criteria: quick wins (simple, repetitive tasks with clear steps) or big time savers (unlock hours weekly with accessible data). Avoid low-priority (complex, low time savings) or deferrable tasks (e.g., quarterly or annual). This systems thinking clarifies inputs, outputs, steps, and criteria, replacing vague departmental views.
Extract Precise Steps and Data with AI Assistance
For each prioritized activity, list explicit steps without vagueness—define terms like "realistic" (e.g., "every phase has 1-week buffer, total length ≤ similar projects"). Use this AI interview prompt to avoid manual overwhelm:
I want you to interview me about a specific process. Dictate/ramble your process here. Ask me one question at a time; each answer informs the next. Uncover every step: what I look at/check, inputs/outputs, vague terms defined. Ask 10-15 questions max. Output: 1) Numbered steps list. 2) Inputs/outputs. 3) Criteria for analysis.
Use fast models (GPT-4o mini, Claude Haiku) for quick back-and-forth. Separately identify inputs (e.g., CSV from project tool, proposal draft) and outputs (e.g., "on/off track" status, approve/edits in specific format). This ensures AI processes exactly what you provide and delivers usable results.
Rank, Prioritize, and Build Focused AI Workflows
Score activities on three axes for starting order: 1) Data readiness (easy to feed AI?), 2) Step clarity (written?), 3) Time savings (hours/week?). Highest scores first. Create one folder per activity on desktop for tools like Claude Co-worker/Code or OpenAI Codex—keeps AI focused for better outputs.
Folder structure (start simple, add complexity later):
- Instructions file:
claude.md(Claude tools) oragents.md(Codex)—paste steps, criteria, rules as persistent prompt. - Input file: Data to process (e.g., proposal draft).
- Output file: AI-generated results.
Scale by client/project: subfolders per engagement (e.g., /clientA/proposal-review). For repeated activities across contexts, bundle into reusable skills (Claude/OpenAI/ChatGPT skills) callable anywhere. This setup turns hours-eating tasks like data extraction into templates, yielding reliable automation from day one.