The Donkey Work AI Delegation Framework
Protect your most human work by deliberately assigning only low-stakes tasks to AI.
This framework gives you a deliberate method for deciding which tasks to hand to AI and which to keep fully human. Every recurring task is categorized as either 'donkey work' (repetitive, administrative, low-stakes) or 'human-critical' (original, identity-defining, high-stakes creative). AI handles the donkey work; you own the human-critical work. This prevents two common failure modes: ignoring AI entirely when it could save real time, and over-delegating in ways that strip creative ownership and require extensive follow-up verification. The result is a workflow where AI genuinely earns its place without undermining authenticity.
- AI can only give you someone else's ideas — protect your original creative voice.
- The real cost of delegation includes the time needed to verify AI outputs.
- Right tool for the right job: sometimes the screwdriver beats the drill.
- Low-stakes, high-volume background tasks are the appropriate domain for AI assistance.
- Keeping humans in control at the most critical decision points preserves quality and identity.
- Audit your full workflowList every recurring task you perform across a typical project or work cycle. Include both creative tasks (ideation, drafting, editing) and administrative ones (scheduling, document tracking, summaries, replies).Pro tipUse a recent completed project as your audit source — walk it back step by step to surface tasks you do automatically.
- Categorize each task as donkey work or human-criticalLabel every task on your list as either 'donkey work' (repetitive, low-stakes, administrative, easily verifiable) or 'human-critical' (original, identity-defining, or requiring judgment that reflects your unique perspective). Ask: 'Would giving this to AI remove something authentic from the output?'Pro tipIf you're unsure, ask whether you'd be embarrassed if your audience knew AI produced that specific element. If yes, it's human-critical.WarningDon't let convenience tempt you into labeling genuinely creative tasks as donkey work — this is how creative voice erodes gradually.
- Assign AI tools to donkey work tasks onlySelect appropriate AI tools or automations for each donkey work category. Common fits include drafting metadata, summarizing transcripts, organizing documents, generating customer support replies, and catching logistical blind spots in plans.Pro tipStart with a single donkey work task and run it for two weeks before expanding — this prevents over-adoption before you've validated fit.
- Establish a written human-critical policyWrite down, as an explicit personal rule, which parts of your workflow will never be delegated to AI. Examples: original video ideas, investment theses, reflective journaling, core creative concepts. Post this where you work.Pro tipPhrase the policy as a positive commitment ('I generate all original ideas myself') rather than a prohibition — it frames the rule as identity rather than restriction.WarningWithout a written policy, the boundary drifts. Convenience is a powerful force.
- Verify all AI outputs before useNever deploy AI output directly. Review every result, edit as needed, and confirm it meets your quality standard. Factor verification time into your estimate of how much the task is actually saving you.Pro tipIf you're spending more than 50% of the original task time verifying, reconsider whether the task was truly donkey work.WarningSkipping verification is where hallucinations, misaligned content, and quality drops enter your workflow unnoticed.
- Reassess your categories quarterlyReview your donkey work and human-critical task lists every three months. New AI capabilities may make previously human-critical tasks delegable, and some donkey work tasks may prove to require more judgment than expected.Pro tipTreat each reassessment as a brief workflow audit — 30 minutes with your task list is enough.
A solo YouTube creator uses ChatGPT at the end of their production process, feeding it the transcript and asking for title and description ideas. They treat these only as inspiration, editing heavily before use. The original video concept, however, always comes from the creator alone — they tried AI-generated video ideas and found they didn't want to make any of them.
A house flipper uses AI to track, organize, and surface the large volume of contracts, permits, inspection reports, and financial records generated by each property deal. The investment strategy, property selection, and pricing decisions remain fully human judgment calls.
A traveler shares a draft Greece itinerary with ChatGPT, providing context about travel style, goals, and motivations. The AI identifies that several planned venues are closed in March, rearranges the itinerary to deliver everything the traveler wanted, and flags seasonal timing issues the traveler hadn't researched. The traveler's preferences and goals drive the entire plan.
Extracted from Mac Power Users podcast, Episode 845, based on a discussion between hosts and guest Patrick Rhone about deliberate and selective AI adoption in personal workflows.