The Art of Delegation to AI:
What to Hand Off and What to Keep

Not everything should be automated. The skill isn't using AI—it's knowing when to use it.

There's a temptation, when you first discover what AI can do, to automate everything. Draft every email. Summarize every document. Generate every first draft. It feels like unlocking a superpower.

Then you realize something uncomfortable: some of those tasks were actually valuable. The thinking that happened while drafting that email? That was where the insight came from. The friction was the feature.

The Delegation Framework

Good managers know that delegation isn't about offloading work—it's about matching tasks to the right capabilities. The same principle applies to AI. Here's a simple framework:

• 01. Delegate to AI: Tasks that are repetitive, rule-based, or require synthesis across large amounts of information
• 02. Keep for yourself: Tasks where the process itself creates value—where thinking through the problem is the point
• 03. Collaborate with AI: Tasks where you need a starting point, but your judgment shapes the final output

The third category is where most knowledge workers should spend their time.

What Could Be Delegated

AI excels at tasks where:

• 01. The output is predictable: Formatting data, converting between formats, following templates
• 02. Speed matters more than nuance: First-pass research, initial drafts, quick summaries
• 03. You're synthesizing information: Combining multiple sources, finding patterns across documents
• 04. The task is interruptible: If you can pause mid-task without losing context, AI can probably do it

These are the tasks that eat your time without feeding your mind.

What Should Stay Human

Some tasks look like good candidates for automation but aren't:

• 01. Strategic decisions: AI can gather information, but the judgment call should be yours
• 02. Relationship-building communication: The personal touch matters; recipients can tell when something is genuine
• 03. Creative problem-solving: The struggle of figuring something out is often where the best ideas emerge
• 04. Learning new domains: You can't outsource understanding

The goal isn't efficiency at all costs. The goal is spending your time on work that matters.

The Collaboration Sweet Spot

The most effective AI usage involves a back-and-forth. You start with your own thinking, use AI to expand or challenge it, then synthesize the result with your judgment.

Example approach: You're preparing a proposal. Instead of asking AI to write it, you write the key points yourself. Then you could ask AI to identify what's missing, suggest counterarguments, or find supporting data. The final proposal would be yours, but it would be better for having used AI as a thinking partner.

The Test

Before delegating a task to AI, ask yourself: "If I never did this task again, would I lose something important?"

If the answer is no—delegate it. If the answer is yes—keep it, or collaborate. The friction might be the point.