Delegation is not dropping an unpleasant task on somebody else. It is not dictating every click, either. You transfer an outcome, enough context, and real decision authority while remaining accountable for the system around the work.
🎙️ Published & recorded: ·
Delegate work when another person can own a meaningful result, when the work develops useful judgment, or when your attention is the bottleneck. Keep decisions whose risk cannot yet be bounded, confidential matters the person cannot legally access, and the manager duties only you can perform. Do not keep work merely because you are faster today. That habit guarantees you will still be the bottleneck next quarter.
# Delegation screen
Outcome: What useful state will exist?
Repeatable: Will this kind of work happen again?
Judgment: Can I explain the decision boundaries?
Access: Can this person receive the required information and tools?
Risk: What is reversible, costly, regulated, or public?
Growth: Will ownership build a capability the team needs?
Delegate: monthly customer-risk review
Keep: final termination decision and private HR records“Can you take this? I'm slammed. It needs to be done by Friday.” The receiver gets urgency without meaning. Repair it in order. One: name the result and why it matters. Two: explain what already happened. Three: state the deadline and what drives it. Four: transfer access. Five: ask the person to play back their understanding before accepting.
The right person has enough baseline skill, room to think, access to the people involved, and a reason to care. The most experienced person is often already overloaded. The newest person may be right when the risk is reversible and support is available. Ask about capacity instead of interpreting silence as agreement. Stretch means one step beyond current judgment, not abandonment in unfamiliar territory.
Candidate Relevant proof Capacity Support needed
Ari ran last review none —
Mei analyzed churn 4 hrs/week first stakeholder call
Jon wants exposure 6 hrs/week data-access approval
# Copyable invitation
“I'd like you to own the July risk review because your churn analysis
showed the judgment this needs. It is about six hours over two weeks.
Before you answer, what would need to move for that to be credible?”Do not call an assignment a development opportunity if saying no carries punishment. Be honest about what is required, what is optional, and what existing work will stop.
A good handoff gives the desired result, reason, customer, relevant history, deadline, constraints, stakeholders, known traps, and source material. Then stop. Context explains the terrain; it does not prescribe every footstep. Ask the owner to propose the route. If you already chose every method, every sentence, and every meeting, you assigned labor but kept the work.
# One-page delegation brief
Result: identify the 10 accounts needing intervention this month
Why: customer-success capacity is limited to 10 recovery plans
Done by: draft Tuesday 15:00; final Thursday 12:00
Use: renewal risk, product usage, open severity-one issues
Do not use: protected demographic data
Stakeholders: Sam in Success; Inez in Data
Constraint: no customer contact before Sam approves the list
History: last list over-weighted ticket volume
Owner proposes: scoring method and review format“Use my spreadsheet, call Sam at nine, sort column G, then send me every email before it goes out.” If those steps are mandatory, explain why. Otherwise fix the handoff. Mark legal and safety constraints. Share the previous file as an example, not scripture. State the result and test. Ask for the owner's approach. Reserve review only for decisions that cross a stated boundary.
“Make it good” is not a standard. Show the audience, decision the work supports, required evidence, acceptable error rate, format, and examples of strong and weak work. Separate must-haves from preferences. If you reveal hidden preferences after delivery, the rework belongs partly to you.
# Definition of done
[ ] exactly 10 accounts, each with a one-sentence risk reason
[ ] source metric and date shown for every claim
[ ] Success and Data disagreements recorded, not silently averaged
[ ] no protected demographic data
[ ] spreadsheet opens without permission requests
[ ] recommendation fits on one screen for Thursday's decision
Preference, not requirement: use the previous color scheme.Use a small sample when quality is hard to describe. Review two completed rows before asking for fifty. A useful example transfers taste faster than a page of adjectives.
Checkpoints exist to catch expensive misunderstanding, unblock decisions, and adjust support. They are not surveillance. Put them after the proposed approach, the first representative sample, and before an irreversible action. Agree on what the owner will bring and which decision the meeting should produce. Between checkpoints, let the person work.
Monday 14:00 — 15 min
Bring: proposed scoring model + two sample accounts
Decide: method is safe enough to run on all accounts
Wednesday 11:00 — async note
Bring: top 10, disputed cases, confidence, help needed
Decide: whether customer-success review can begin
Escalate immediately if: protected data appears, scope exceeds 6 hours,
or the Thursday decision is likely to slip.“Just send me an update every hour so I know you're on track.” Hourly reports steal attention and still do not define track. Replace them. Identify the next costly assumption. Ask for evidence at the point it can be tested. Define urgent escalation triggers. Cancel routine checks once the owner demonstrates control. Anxiety is real, but reporting theater is not the cure.
Reverse delegation happens when an employee brings a problem and the manager quietly takes ownership back. Do not answer every uncertain question with “I'll handle it.” Keep the owner while supplying judgment. Ask what they observed, what options they considered, what they recommend, and what specific help they need. Sometimes the honest answer is that your missing decision blocks them; make it promptly.
# When someone says “What should I do?”
“What decision are you trying to make?”
“What are the two best options you see?”
“What do you recommend, and what risk worries you?”
“Which part needs my authority or information?”
# Keep ownership explicit
“You still own the account list. I will get Data to restore your access
by noon. If it is not restored, bring me a reduced-scope recommendation.”“The stakeholders disagree, so can you just finish the deck?” Fix the blockage, not the whole assignment. Ask the owner to write each disagreement. Identify which one requires your authority. Make or escalate only that decision. Return the next action and deadline to the owner. If the brief was defective, admit it and repair the brief instead of calling the employee unready.
Review against the agreed result, not against the version you happened to imagine later. Point to the observable gap, its effect, and the standard. Ask what led to the choice before prescribing a correction. For rework, say what must change, what can stay, who owns it, and when it will be checked. Praise judgment specifically so it can be repeated.
# Feedback script
“The source dates are missing on 7 of 10 rows. That means Thursday's
reviewers cannot tell whether the risk is current. Our done checklist
requires a metric and date for every claim. What made those dates hard
to include? Please keep the account choices, add the dates, and flag any
source older than 30 days by 10 tomorrow. You still own the final.”
# Specific praise
“You recorded the disagreement instead of averaging it away. That gave
us a real decision to make. Use that move again.”If the work met every stated standard but you dislike its style, call that a preference and decide whether changing it is worth the cost. Do not rewrite the work in secret. That teaches the owner to wait for your invisible answer key.
Delegation needs room for ordinary mistakes and hard stops for dangerous ones. Name reversible experiments, spending limits, privacy rules, public commitments, production access, and the signal to stop. In remote work, write the brief, authority, decisions, and checkpoint times in one durable place. Time-zone silence is not consent. State who can decide when the manager is asleep.
# Safe operating boundary
You may: test scoring on internal data; revise weights; consult owners
Ask first: contact customers; add a data source; spend over $250
Never: export personal data; paste customer records into public tools
Stop and page: suspected data exposure or wrong-account contact
If I am offline: Priya may approve scope changes up to one day
Decision log: date · choice · owner · evidence · people informed“I posted it in chat yesterday and nobody objected, so I assumed it was approved.” Repair the process. One: pause the action if it can still be reversed. Two: identify the named approver. Three: get an explicit yes or no. Four: record the decision in the shared brief. Five: add a response deadline and backup approver for the next async decision.
The shortest useful delegation conversation covers outcome, reason, owner, authority, definition of done, constraints, resources, checkpoints, escalation triggers, and playback. Put it in writing for work that spans days or people. Then judge the system as well as the person when it goes wrong.
# Copy, fill, send
OUTCOME:
WHY / WHO USES IT:
OWNER:
DONE BY:
DONE MEANS:
BACKGROUND / SOURCES:
AUTHORITY LEVEL: 1 research · 2 recommend · 3 consult+decide
4 decide+inform · 5 independent in bounds
MUST / MUST NOT:
RESOURCES AND ACCESS:
CHECKPOINTS + DECISIONS:
ESCALATE WHEN:
BACKUP APPROVER:
PLAYBACK: “Tell me your understanding and first move.”
# If delivery fails
protect impact → compare with stated done → ask what happened
→ own briefing/access failures → correct the gap → return ownership
→ change the system that allowed recurrenceFinal position: delegation is neither task disposal nor remote control. Give a person a result worth owning, the authority to make real choices, and boundaries that make mistakes affordable. Keep accountability. Give up the fantasy that your exact method is the only competent one.