Delegation Decision Checklist for Everyday Tasks
Use this checklist before starting a task to decide whether to do it yourself, hand it to a teammate, or delegate it to Claude.
It works as a quick verification pass: walk the questions in order, and let your answers point you toward the right owner for this specific task.
How to Use This Checklist
- Run through the checks for one task at a time; the right owner can differ even between two tasks that look similar on the surface.
- A "no" on a check is not automatically disqualifying, it is a signal to look closer, and it often points toward partial delegation rather than a flat no.
- Revisit the checklist if a task's stakes or audience change partway through, since that can flip the answer.
- Keep the interpretation guide at the end handy the first several times you use this; the pattern becomes intuitive with repetition.
Task Shape
- Does the task have a clear input and a recognizable "done"? Bounded tasks, like summarizing a document or drafting an outline, are easier to hand off than open-ended ones.
- Is this task repetitive or similar to something done many times before? Repetitive drafting, formatting, and research synthesis are strong Claude candidates.
- Does the task require deep familiarity with unwritten context, like team politics or history not documented anywhere? If yes, a teammate who holds that context may be a better fit than either DIY or Claude.
- Is the task mostly first-pass work that will be reviewed and revised afterward? First drafts, initial research passes, and rough outlines suit delegation better than final, polished output.
Stakes and Reversibility
- If the output is wrong, is the mistake easy and cheap to catch before it matters? Low-cost, reversible mistakes make a task safer to delegate.
- Does this task involve an action that cannot be undone once taken, like sending, filing, or publishing? Irreversible actions call for more human involvement before anything goes out.
- Is this a one-off high-stakes deliverable, or one of many lower-stakes items? Single high-stakes items deserve more direct human attention than volume work.
- Would a wrong or off-target result damage trust with a customer, employee, or partner if it slipped through? Higher relationship risk argues for keeping a human closer to the output.
Accountability
- Does this task end in a final judgment call, such as a hiring, firing, legal, or financial decision? Final judgment calls on matters like these should stay human-led, even if drafting the resulting communication can be delegated.
- Can you clearly say who is accountable for this outcome if someone asks later? If the honest answer has to be a person, the decision itself should be made by that person, not generated.
- Does the task affect someone's livelihood, safety, or legal standing? Tasks with this kind of impact belong in the human-led or partial-delegation category, not full delegation.
- Is there a human checkpoint planned before this output goes anywhere that matters? A planned review step is one of the things that makes delegation safe in the first place.
Fit for a Teammate vs. Claude
- Does the task need specialized institutional knowledge a teammate has and Claude does not, like unwritten relationship history? That points to a teammate over Claude.
- Is the task mostly volume, pattern-finding, or drafting work across a lot of similar material? That points to Claude over a teammate, since this is where synthesis at scale helps most.
- Would delegating to a teammate create a bottleneck (their availability, workload) that this task cannot wait for? A bottlenecked teammate task may be a better fit for Claude's first pass, with the teammate reviewing instead of originating.
- Can you write a description with enough context and constraints for the task to make sense to someone without your full background? If you cannot describe it clearly enough for anyone else, human or Claude, to act on, the task may need more thinking before it is handed off to anyone.
Interpreting Your Answers
- Mostly "yes" across Task Shape, Stakes and Reversibility, and Accountability: this is a strong candidate for full delegation to Claude.
- "No" on any Accountability item: keep the decision itself human-led; consider delegating only the drafting or research that follows once the decision is made.
- "No" on Task Shape items about unwritten context: consider a teammate for the parts that need that context, and Claude for the parts that do not.
- Mixed answers across a single task: split the task at the seam between its decision and its drafting, and route each half separately rather than forcing one owner for the whole thing.
FAQs
What is this checklist for?
Deciding, task by task, whether to do something yourself, hand it to a teammate, or delegate it to Claude, using the same set of checks each time.
What if a task fails one check but passes most others?
A single "no" is a signal to look closer, not an automatic disqualification; it often points toward partial delegation, keeping the decision-making part human while delegating the drafting or research part.
Why does the checklist separate "Stakes and Reversibility" from "Accountability"?
- Stakes and reversibility are about how costly a mistake would be and how easy it is to catch.
- Accountability is about who owns the outcome and must be able to explain the decision, which is a separate question even for a low-stakes-seeming task.
How do I know if a task needs a teammate instead of Claude?
If it depends on unwritten institutional knowledge, like team history or relationship context that exists only in people's heads, a teammate who holds that context is usually a better fit than Claude.
Can a task be split between a teammate, Claude, and yourself?
Yes. Splitting at the seam between decision and drafting, or between context-dependent and pattern-finding portions, is common and often produces a better result than forcing one owner for the whole task.
What is the single most important question on this checklist?
Whether you can clearly say who is accountable for the outcome; if the honest answer has to be a specific person, the decision itself belongs to that person, no matter how the rest of the checklist scores.
Does a task with high volume always suit Claude?
Usually, because per-item stakes tend to be lower and errors are easier to catch in aggregate, but a high-volume task that touches sensitive individual cases, like performance reviews, can still fail the Accountability checks and need more human involvement.
Should I run this checklist for every single task, even small ones?
For small, low-stakes, routine tasks, the checklist becomes close to instant with practice; for anything unfamiliar, high-stakes, or ambiguous, running through it deliberately is worth the extra minute.
What should I do if the checklist says a task is a strong delegation candidate?
Move on to writing a clear description, context, constraints, and the actual goal, so Claude can act on the delegation decision you just made.
What should I do if my answers keep changing for the same type of task?
Treat that as a sign the task type is context-dependent rather than fixed, and keep running the checklist per instance instead of assuming a past answer still applies.
Does passing every check guarantee the output will be good?
No, the checklist only decides who should attempt the task first; reviewing the resulting output before relying on it is still necessary regardless of who or what produced it.
Related
- Delegation: Deciding What to Hand Off to Claude - the underlying concepts behind these checks.
- AI Fluency Framework Basics - worked examples of these decisions in practice.
- Description: Crafting Clear Instructions for Claude - what to do once the checklist points to Claude.
- Delegation & Description Best Practices - broader habits that build on this checklist.
Stack versions: Written against the Claude model lineup current as of ~June 2026 - Claude Fable 5, Claude Opus 4.8, Claude Sonnet 5 (the default), and Claude Haiku 4.5. Model names, pricing, and product features move quickly - verify current specifics at platform.claude.com/docs before relying on them.