Delegation: What to Hand Off

Which tasks are ready to hand off

Some work is a natural fit for an AI employee right away. A task you would describe as repetitive, rules-based, or heavy on text and data is what people call an automation candidate: it has a clear input, a clear output, and not much hidden judgment in between. Drafting a reply from a template, sorting a list, pulling figures into a summary, and turning notes into a first draft all qualify. Other work is not ready, at least not yet. Anything that leans on taste, a relationship, or a high-stakes call belongs with you for now: deciding who to hire, having a hard conversation, signing off on a contract. A good rule is to start by handing off the tasks you would happily explain to a competent new colleague, and keep the ones you would not.

• Ready: repetitive, rules-based, text or data heavy, clear input and output.
• Keep for now: real judgment, relationships, anything high-stakes or one-way.

Treat it like a capable new hire

The skill that makes this work is plain delegation, the same thing a good manager does with a person. You give three things: the goal you want reached, the context it needs to reach it, and the standard the result has to meet. You do not narrate every click, because a capable worker should choose the steps. This is where most people go wrong. They either say too little ("write me something") and get a generic result, or they micromanage the exact steps and lose the point of delegating at all. Aim for the middle: be clear about the outcome and the bar, hand over the background, then let it do the work and review what comes back.

• Goal: what done looks like.
• Context: the background, examples, and facts it cannot guess.
• Standard: tone, length, format, and what good enough means.

Knowing when to step in

Delegating does not mean walking away. For anything irreversible, sensitive, or expensive, you stay involved and approve the result before it goes out. The name for keeping a person at the decision point is human-in-the-loop: the AI employee does the work, and you sign off where it matters. For everything else, let it run. Checking every small, low-risk task by hand defeats the purpose and wears you out. Match the oversight to the stakes: review closely where a mistake would be hard to undo, and trust the AI employee to handle the routine on its own.

• Stay in the loop: irreversible, sensitive, or high-cost actions.
• Let it run: routine, low-risk, easy-to-undo work.

Key takeaways

• Start with automation candidates: repetitive, rules-based, text and data work.
• Keep judgment, relationship, and high-stakes calls with yourself for now.
• Delegate like a manager: give the goal, the context, and the standard, not the clicks.
• Stay human-in-the-loop for anything irreversible or sensitive, and let the rest run.

Continue learning

• Working With AI Employees — View the complete course.
• Next: Description: Asking for Great Results