A leader becomes a decision bottleneck when work repeatedly waits for that person's answer and the available review capacity cannot keep pace. The answer is not automatically to delegate more or remove approvals. First identify which decisions are waiting, why they require that leader and what happens before they reach the queue.
A decision-queue audit is a bounded observation of work, not a personality assessment. It helps distinguish essential leadership judgment from unclear authority, incomplete requests and inherited approval habits. The accountability versus control guide explains the broader leadership tension; this resource offers a practical way to investigate the waiting.
Choose a defined slice of decisions
Select a work area and a representative observation period. Include different types of decisions rather than only a dramatic example that confirms what the team already believes. Use approved records and avoid copying confidential personnel, customer or financial details into an informal log.
Capture the decision requested, arrival point, authorized decision owner, information available and time of response. Note whether the request was actionable when it arrived. A queue with incomplete requests is different from a queue of ready decisions waiting for scarce judgment.
Working tool (AI-assisted editorial): Download the decision queue aging workbook to keep original arrival, readiness and closure separate. Record one distinct decision per ID; use its calendar-day ages and checks for missing owners, conflicting dates or duplicate IDs when reviewing the queue. A follow-up message should not restart the original arrival clock. The workbook helps organize the observation; it does not set an approval deadline or change decision authority.
Classify why each decision waits
Useful categories include missing information, unclear authority, competing priorities, unavailable reviewer and a required approval that is functioning as designed. Treat the categories as provisional explanations, not confirmed causes. One request can pass through more than one condition.
Ask the people sending and receiving the request to explain the same example. The sender may believe a complete question was submitted while the reviewer sees an unsupported recommendation. That disagreement is part of the work to clarify, not proof that either side is careless.
Distinguish essential judgment from inherited habit
Some approvals exist for legitimate risk, policy or contractual reasons. Keep them governed by the appropriate authorized owners. For other decisions, ask why the leader is involved and whether the reason is current, explicit and supported.
A hypothetical manager might discover that routine schedule adjustments come to them because nobody knows the supervisor's boundary. Another request may genuinely require a senior priority decision. These should not receive the same response: clarify the first boundary through the authorized process and preserve the second decision with a usable review route.
Improve the request before expanding the queue
Define the minimum useful input for a decision: the choice required, relevant evidence, options, recommendation where appropriate, unresolved uncertainty and needed response point. This is not an invitation to create a large form for every minor question. Match the effort to the consequence.
Test the agreement on a small number of suitable requests. Did the reviewer still need to ask for missing facts? Did the sender understand what was required? Better request quality can release review capacity without transferring authority or weakening necessary oversight.
Clarify decision rights through an authorized agreement
Use the decision rights matrix example to name who decides, whose input is needed and where escalation applies. Separate preparing the recommendation from approving the decision. Being capable of drafting an answer does not establish authority to act on it.
Where a boundary changes, have the appropriate owner approve it and make the change understandable to receiving teams. Ask a supervisor to explain what they may decide and which condition sends the matter onward. Do not rely on a general message to use judgment when the boundary is what created the queue.
Are repeated messages one decision or several?
Queue-counting example (AI-assisted editorial; fictional): A supervisor sends a decision request, follows up by email and raises it again in the leadership meeting. Those contacts may concern one unchanged choice rather than three separate decisions. Compare the work reference, requested choice, current evidence and authorized owner before counting them. Preserve the contact history and response needs; do not delete reminders simply to make the queue look smaller.
Keep a decision’s arrival and readiness history visible when another message is added. A follow-up should not reset the apparent age of an unanswered, actionable request. Conversely, a materially changed choice or a separate required approval may need its own entry even if it concerns the same project. Record the relationship rather than merging it automatically. Review both the number of distinct decisions waiting and the clarification or follow-up effort they generate. This example helps define the observation, not remove approvals, assign blame or establish a universal response target.
Examine review capacity as well as authority
If appropriate decisions still exceed the leader's available time, review the volume, consequence and cadence. Could ready decisions be reviewed at a predictable point? Are similar requests repeatedly reopened because the decision was not recorded? Does the leader need qualified support rather than simply a faster response target?
A review slot must not become permission to wait through urgent conditions that have an established escalation procedure. Keep those procedures intact. The useful change may be better preparation, more appropriate review capacity or a leadership choice to reduce competing work.
What does average response time leave out?
Response-time example (AI-assisted editorial; fictional): Five completed requests needed one, one, one, one and six calendar days from the same defined readiness point to an authorized decision. A follow-up message does not restart the wait. Their mean is two days: ten days divided by five requests. Four were answered within two days; one took six. The two-day mean therefore does not mean every request was answered within two days. This constructed sample describes those five completed requests, not the likelihood or promised timing of future decisions.
The Kanban Guide (May 2025) describes a service level expectation as a forecast with an elapsed time and probability. That differs from a required deadline, and this five-request illustration is not a reliable future forecast. Review individual waits and currently unanswered requests alongside the mean. A long wait can involve missing evidence, required specialist review or unavailable capacity; the duration alone does not identify its cause. Check the applicable response need and authorized owner before choosing a repair. Preserve required approvals.
Test the whole flow after the change
Compare similar decisions before and after, using consistent definitions. Review waiting, clarification, rework and whether the authorized owner is still making the appropriate decisions. Faster responses are not an improvement if they create more reversals or unsupported commitments downstream.
Bring recurring priority conflicts to the weekly leadership review. If the queue reflects a wider leadership-system issue, discuss the situation with Shannon. An effective audit makes the next operating decision clearer; it does not merely label the leader as the problem.
About this resource: Prepared with AI-assisted drafting from published guidance. Examples are illustrative, not reported client results. Cover illustration: AI-generated.