People Management
Delegate a Decision, Not Just a Task: Agree on Authority and Checkpoints
A completed runbook-review delegation and an approval-heavy comparison show what the employee may decide, when to consult, and when the manager must act.
To delegate a decision, state what the other person may decide without another approval, what input they must seek, and which conditions require escalation. Agree on the result, constraints, and checkpoints before work starts. The manager still sets priorities, provides resources, and addresses problems. Do not hold someone responsible for a decision you required them to leave with you.
“Please handle it” leaves room for very different expectations. One person hears permission to act; another hears a request to prepare a recommendation. This guide from LandOffer.ai uses a fictional engineering-management case to make that distinction concrete. The examples do not grant real financial, personnel, security, or operational authority.

Name the decision instead of assigning an ambiguous outcome
Write the choice as a question: “Which review method will we use for these twelve runbooks during the next four weeks?” That gives the delegate something to decide. “Improve our documentation” could mean prioritizing pages, changing procedures, buying a tool, or running a review. Those choices may belong to different people.
Identify the consequence and the boundary. A team member can choose a review method without being authorized to change a production procedure. They can organize work within agreed capacity without committing another team’s time. Narrowing the decision is especially useful when a larger project contains several owners.
OpenStax’s discussion of organizational authority distinguishes delegated authority and responsibility from accountability that continues through the management hierarchy. This is a management principle, not a determination of legal liability in your organization. Use it to ask whether the person receiving the work has meaningful choice and whether the manager’s continuing obligations are clear.
If you want only a recommendation, say that. “Compare the two methods and recommend one; I will decide Friday” is a legitimate assignment. It is different from “Choose the method within these constraints and tell me your reasoning.” Confusion follows when you promise decision authority but continue making the choice yourself.
Separate deciding, consulting, and escalating
For a bounded delegation, define three kinds of action. The delegate decides inside the agreed scope. They consult named people for necessary knowledge or effects on their work. They escalate when a constraint is breached or a decision outside their authority is required. These are practical distinctions, not a universal set of formal delegation levels.
Use precise verbs. MIT’s job-description glossary distinguishes coordinating work from authorizing action and exercising final authority. Although that resource concerns position descriptions, the language is useful here: asking someone to coordinate several contributors does not automatically make them the final decision-maker. A verb such as “own” needs an explanation of the actual rights attached to it.
Consultation should produce information the delegate can use. State whose input is required, what question they should answer, and when it is needed. Do not imply that every contributor has a veto unless that is the actual arrangement. Where a formal policy requires approval, preserve it explicitly rather than renaming it consultation.
Escalation needs a destination and a trigger. “Use your judgment” is incomplete if the employee does not know what to do when a choice affects security, spending, another team, or the agreed scope. Identify which already-authorized work can continue while the exception is reviewed, so escalation does not leave the employee guessing whether everything must stop. A manager’s absence should not silently turn missing authority into permission.
Completed delegation agreement: choose a runbook-review method
The following is a fictional teaching case. All names, dates, documents, hours, and outcomes are invented. No internal documents were reviewed, tools tested, or workplace messages sent.
Engineering manager Nadia asks Owen, a senior engineer, to choose a review method for twelve priority runbooks during an October 5–30, 2026 pilot. The aim is to identify stale links and route possible content problems to the authorized runbook owners. It is not permission to edit operational procedures or change production systems. The example assumes Nadia has authority to delegate this narrow workflow choice.
| Agreement field | Completed fictional arrangement |
|---|---|
| Decision and deadline | Owen selects the review method for the twelve listed runbooks by October 9 and records the reasoning |
| Owen may decide | How to combine an already-approved internal link checker with manual review, and how to sequence the twelve runbooks within the pilot |
| Required consultation | Runbook owners identify critical sections and confirm where suspected content issues should go; the internal tooling owner explains the approved checker’s limits |
| Constraints | At most eight total team hours, including consultation and review. Use approved internal systems only; no new spending, external document upload, or changes to operational procedures |
| Escalation triggers | Expected effort exceeds eight hours; an unapproved tool is needed; a security concern appears; or the proposal requires another team’s commitment or a procedure change |
| Checkpoints | October 7 scope and evidence check before the choice; October 23 progress and risk check. Neither is an extra method-approval gate if the decision stays within scope |
| Communication and record | Owen records the selected method, consulted input, trade-offs, and unresolved issues; informs Nadia and the affected owners after deciding |
| Nadia retains | Protecting the agreed capacity, resolving conflicting priorities, arranging help, confirming the authority boundary, and reviewing the pilot’s result |
| Closeout | October 30 review of completed checks and unresolved items; owners receive the issue list, and any continuing review process needs an explicit decision |
Owen has a real choice: a manual review, the approved checker plus targeted manual review, or another combination within the stated boundary. He does not need Nadia to approve the same choice again merely because she would have preferred a different sequence. If Nadia actually needs to retain final approval, the agreement should say so before Owen begins.
The runbook owners also retain meaningful responsibilities. Their input can reveal that a page has a critical section or a procedure problem. Owen cannot convert permission to identify a problem into permission to alter instructions used in operations. The scope protects the distinction between finding an issue and approving its remedy.
Choose checkpoints that examine risk rather than every small action
A checkpoint should answer a useful question. Before an unfamiliar decision, check whether the delegate understands the constraints and has access to the needed information. During the work, check progress, new risks, and support. Afterward, examine the reasoning and result. The timing depends on the decision’s uncertainty, consequence, and reversibility.
Atlassian’s DACI guidance distinguishes the person who drives a decision process from the person who makes the decision, and from people providing expertise or receiving the outcome. That is a useful role check, not a requirement to adopt a particular tool or label. Calling Owen the driver while Nadia remains the decision-maker would describe a different arrangement from this example’s delegated choice.
More frequent review can be appropriate for a new kind of work or a more consequential choice. Explain the reason and what will change as experience develops. If every email, question, and draft must wait for the manager, the employee may be carrying coordination work while the decision remains centralized. That can be an intentional arrangement, but it should not be advertised as autonomy.
For Owen’s bounded pilot, the October 7 conversation examines scope, capacity, and access before the October 9 choice. The October 23 conversation examines progress and exceptions. Nadia can ask why Owen rejected an option without reopening the decision solely because his preference differs from hers. A new constraint is a reason to reconsider; surprise at a reasonable choice suggests the original boundary needed clarification.
Worked comparison: approval bottlenecks versus bounded choice
Continue the fictional case with two alternative versions of the same assignment. Neither version is an observed workplace test, and the comparison establishes no measured productivity improvement.
In the first version, Nadia says “You decide,” but asks Owen to obtain approval before contacting each owner, before choosing the sample order, before running the approved checker, before writing the issue list, and before communicating the method. Owen prepares material at each step while Nadia makes the practical choices. The repeated waits are part of this invented setup, not a quantified claim about how all managers behave.
In the revised version, Nadia and Owen use the filled agreement. Owen consults the tooling owner, who confirms that the approved checker detects broken internal links but does not assess procedural accuracy. Runbook owners identify the sections that need careful human reading. Owen selects the checker plus targeted manual review and records why: the checker addresses link status, while the manual pass identifies possible content concerns for the proper owners.
| Situation | Approval-heavy version | Bounded-decision version |
|---|---|---|
| Ask a listed owner about a critical section | Wait for Nadia to approve the contact | Owen seeks the required input directly |
| Sequence the twelve runbooks | Nadia approves the order | Owen chooses within the agreed pilot and explains the priority |
| Use the existing approved checker | Ask again despite unchanged scope | Owen uses it within its documented limits |
| Consider a new external crawler | Unclear because “you decide” has no boundary | Escalate before sharing internal URLs or documents |
| Expected effort reaches ten hours | Continue and explain later, or wait without a rule | Raise the two-hour gap before exceeding the eight-hour allocation |
The external-crawler branch changes the decision. Owen cannot interpret faster checking as permission to send internal material outside approved systems. He reports the limitation and asks Nadia to choose a permitted response: reduce scope, provide more approved support, or seek the necessary review. Until then, he continues only the portion already authorized and feasible.
The capacity branch also has a concrete consequence. Ten expected hours against eight approved hours leaves a two-hour gap. Owen brings the work estimate and options; he does not independently take another team’s time or assume the extra effort is free. Nadia’s next responsibility is to resolve the trade-off; raising an agreed trigger is part of Owen’s role.

Respond to an escalation without taking back every decision
When someone escalates, first identify whether the trigger occurred. Ask what changed, which constraint is affected, what evidence supports the concern, and which options remain. Respond to the exception they raised. A request for two additional hours does not necessarily require rewriting their whole method or withdrawing every delegated choice.
In the fictional ten-hour branch, Nadia might narrow the pilot from twelve runbooks to ten, or obtain an approved capacity change. Either would alter the agreement and should be recorded after checking that the revised scope or capacity is feasible and properly approved. The example does not claim one is always better; the manager needs the actual risk and priority information. Owen then proceeds within the revised boundary.
If the delegate makes a poor choice inside a genuinely authorized scope, review the reasoning, missing information, and consequences. Do not retrospectively claim they lacked authority simply because the result disappointed you. If a necessary control was omitted, address that control. If support was unavailable, acknowledge the manager’s part in the situation. Accountability can involve several people with different responsibilities.
Likewise, a delegate should not hide a breached constraint behind “You told me to decide.” Decision rights are bounded. Preserve both sides of the agreement: meaningful discretion within the limit, and prompt escalation when the limit is reached.
Close the loop and adjust the next delegation
At closeout, compare the outcome with the original purpose. Which runbooks were checked, which issues were handed to owners, and what remains unresolved? Keep link checking separate from content approval. A completed scan does not establish that every procedure is safe, current, or correct.
Discuss how Owen used the owners’ input, compared the review methods, and raised the capacity gap, as well as what the pilot produced. Did Owen collect relevant input, explain trade-offs, stay inside the constraints, and escalate the exception when it became visible? If future delegation expands, identify the new scope and support instead of assuming this pilot authorized all documentation or operational decisions.
Start with one live management choice: write what the employee can decide, whose input is required, and the first condition that sends it back to you. Confirm those three points with the delegate and the required contributors before the decision work begins. LandOffer.ai publishes this management guidance; the actual organization’s authority and control requirements still govern the work.


