How to Decide What Work AI Should Do for You: The AI Deputisation Audit

Key Takeaways

• AI Deputisation is a method for deciding what work to hand to AI, rather than simply looking for things to automate. Start with a recurring task, assess it across five dimensions, then decide whether to deputise it to AI, work on it as a duet with AI, or defend it as primarily human work.
• 1. Time — How much of your time does the task consume? Start with work that happens repeatedly and absorbs meaningful time; the larger the recurring time cost, the greater the potential value of delegating it to AI.
• 2. Teachability — How easily can you show AI how the task is done? A task is a stronger candidate when its process can be explained, documented, or demonstrated step by step, including through emerging “teach-a-task” approaches.
• 3. Checkability — How easily can you tell whether AI did the task correctly? Delegation becomes much safer when reviewing the output takes substantially less effort than doing the work yourself and when success or failure is relatively easy to recognise.
• 4. Error cost — What happens if AI gets it wrong and you do not notice? Tasks with reversible, low-impact mistakes are better early candidates for deputization; tasks where an unnoticed error could cause serious financial, reputational, legal, or operational damage require much greater human involvement.
• 5. You-dependence — How much does good performance depend specifically on your judgement, taste, relationships, or experience? The more the value of the task comes from something uniquely human or uniquely yours, the stronger the case for keeping yourself in the loop rather than fully handing it over to AI.