Which Tasks Are Actually Worth Handing to Claude?
New users tend to land on one of two habits, and both waste time. The first only reaches for Claude on the rare occasion they remember it exists, missing dozens of smaller opportunities every week. The second reaches for it on everything, including tasks where writing a good prompt takes longer than just doing the task by hand. Neither is a Claude problem. Both come from never having a quick way to tell a good candidate from a bad one.
Tip
The question isn't "can Claude do this." It's "does drafting a version and reviewing it actually beat doing this myself."
A task worth handing off isn't the hardest one on your list. It's the one where a rough first pass saves more time than it costs to review.
A three-part test
- 1
Is it repetitive or time-consuming enough to matter
A task you do once a year, taking ten minutes, isn't worth the setup of explaining full context to get a first draft. A task you do every week, or one that would normally eat an hour, is.
- 2
Is a rough first pass low-risk to get wrong
If reviewing a draft and fixing what's off is faster than writing from a blank page, this is a good candidate. If a wrong first draft could cause real damage before anyone catches it, that's a signal to draft with much tighter constraints, or not to draft at all.
- 3
Does it depend on knowledge only you actually have
A task that hinges on office politics, an unstated relationship history, or context that exists only in your head isn't a good fit until you can actually put that context into words. If you can't explain it to Claude, it can't factor into the draft.
A task that clears all three is worth handing off. One that fails even one is either not worth the setup, too risky to draft first, or missing information you'd need to supply anyway.
I write a weekly status update for my team, roughly 200 words, same structure every time: wins, blockers, next week's priorities. Here are my raw notes from this week. Draft the update in that structure.
”That clears the test cleanly: it's weekly and repetitive, a rough draft is easy to fix and low-risk, and nothing about it depends on hidden context you can't write down.
What a bad candidate actually looks like
This fails the third part of the test outright. A promotion decision depends on performance history, team dynamics, and organizational context that exists in your head and your manager's, not in a prompt. Claude can help you organize the factors you're weighing once you've laid them out, but it can't supply the judgment call itself, because the actual deciding information was never handed over and probably shouldn't be typed into a chat in the first place.
The distinction matters: this isn't a task Claude is bad at in some absolute sense. It's a task where the useful part, structuring your own reasoning, is a smaller piece than the part Claude can't touch at all.
Inside Claude Tutorial
This same filter gets faster with practice.
Recognizing a good candidate becomes automatic once you've run the test on enough real tasks. The app builds this judgment through short, repeated practice rather than a single explanation.
An edge case: tasks that start bad and become good
Some tasks fail the test today only because the missing piece hasn't been written down yet. A recurring client email that "depends on knowing the relationship" becomes a good candidate the moment you write two sentences summarizing that relationship once. The test isn't a permanent verdict on a type of task. It's a check on whether everything the task actually depends on can currently be put into words. Once it can, a bad candidate becomes a good one.
Continue reading
- Getting Started with Claude: the foundational habits this test builds on.
- A Daily Workflow for Using Claude Well: applying this same test to build a sustainable daily routine, not just a one-off decision.