This builds on my recent piece on decision rights in autonomous teams, because visibility and restraint only work when it is clear who decides what, and how misalignment escalates.

‍

Leaders in modern product and engineering organisations live in a tension that never fully goes away. If you stay too visible, teams become dependent, decisions slow down, and accountability drifts upward. If you stay too restrained, teams fill the vacuum with local optimisation, and misalignment becomes expensive.

‍

This is not a personality issue. It is an operating model issue. Visibility works when it is a service. Restraint works when decision rights are explicit.

‍

A useful starting point is intent. Mission command doctrine defines a model where leaders set intent and boundaries, then enable disciplined initiative (Perna, 2019). In business terms, leaders should be present enough to sharpen intent, reduce ambiguity, and remove blockers. Leaders should be restrained enough to keep ownership where the work lives.

‍

What this tension looks like in the real world

1. Leaders step in late. They discover that a “small” local decision has created cross-team coupling, or undermined an SLA, or increased risk.
2. Leaders step in early and often. Teams stop proposing options and start waiting for answers.
3. Teams interpret visibility as mistrust. Leaders interpret restraint as non-compliance.
4. Decisions repeat. Teams re-litigate choices because nobody knows who can decide, and what is non-negotiable.

‍

Rogers and Blenko describe this as decision bottlenecks created by unclear roles, especially across functions, business units, and partners (Rogers & Blenko, 2006).

‍

Why this happens

1. The organisation confuses autonomy with absence. Autonomy still needs guardrails.
2. Decision rights are implicit. People cannot answer, “Who decides,” without debate.
3. Leaders use visibility as control. It feels faster in the moment, but it creates learned helplessness.
4. Leaders use restraint as avoidance. If conflict feels costly, leaders may soften direction, then get surprised later.
5. Psychological safety is uneven, including among leaders. When leaders do not feel safe to challenge peers or escalate misalignment, they accept ambiguous commitments. Teams then optimise locally (Edmondson, 1999).

‍

Marquet’s warning lands here. “Control without competence is chaos” (Marquet, 2013). The inverse is also true. Competence without clarity becomes fragmentation.

‍

A practical frame for “visibility as a service”

‍

You can run visibility and restraint as three deliberate modes. This keeps you consistent, and it stops your team from guessing what you mean by autonomy.

‍

‍

‍Remote control anti-pattern: You are not present for context, but you still override decisions through escalations or last-minute directives. This destroys trust and ownership.

‍

Mode 1. Clarity mode. High visibility, low control

‍Use this when you need shared intent.

‍

Triggers:

1. Cross-team impact.
2. Policy or risk posture changes.
3. Any decision that is hard to reverse.

‍

Leader behaviours:

1. State the outcome and why it matters.
2. Define non-negotiables and decision boundaries.
3. Assign a single owner for the decision and the outcome.

‍

Mechanism: Write it down. A one-page decision packet is enough.

‍

Mode 2. Coaching mode. Medium visibility, medium guidance

‍Use this when the team owns the decision but needs support to reason through trade-offs.

‍

Triggers:

1. New domain.
2. Repeated rework.
3. Capability gaps.

‍

Leader behaviours:

1. Ask for options, risks, and measures.
2. Challenge assumptions without providing the solution.
3. Time-box discussion, then commit.

‍

A useful phrase from Amazon’s shareholder letter is “disagree and commit” (Bezos, 2016). It reduces circular debate, but it only works if decision rights are clear.

‍

Mode 3. Intervention mode. High visibility, high control

‍Use this sparingly. It is sometimes necessary, but it has a cost.

‍

Triggers:

1. Safety or compliance breach.
2. An SLA breach trend with no corrective action.
3. A decision that violates published non-negotiables.

‍

Leader behaviours:

1. Take the decision, name it explicitly, and document why.
2. Stabilise the system.
3. After stabilisation, return ownership to the team with clearer guardrails.

‍

Mode 4. Remote control. Anti-pattern. Low visibility, high control

‍This is the failure mode where a leader is largely absent from context-building, but still overrides decisions late through escalations, side channels, or last-minute directives. It feels efficient in the moment. It reliably destroys ownership.

‍

Common triggers:

1. Leaders are overloaded and only engage at the last responsible moment.
2. Decision rights are unclear, so leaders intervene to prevent perceived risk.
3. Stakeholders escalate directly to leaders, bypassing the agreed decision path.

‍

What it causes:

1. Teams stop bringing options early, because they expect late overrides.
2. Information quality drops, because decisions move to informal channels.
3. Accountability becomes performative. Delivery owns outcomes, but not the decision.

‍

How to correct it:

1. Move the override earlier. If you must be involved, switch to Clarity mode and set intent and boundaries upfront.
2. Create a documented escalation path. Escalate to the decision forum, not to the person.
3. Use a decision packet for one-way-door decisions, so leadership can engage without taking execution ownership.
4. Name a Directly Responsible Individual and keep the “D” visible, so overrides become rare exceptions rather than a pattern.

‍

This is the leadership version of servant-first thinking. “The servant-leader is servant first” (Greenleaf, 1970). Visibility serves clarity and capability. Restraint serves ownership.

‍

A consultancy example that shows the same dynamic

‍

Scenario: A consultancy team recommends a safer rollout approach that protects uptime and reduces rollback risk. The client sponsor chooses the faster shortcut, without fully understanding the implications.

‍

If the consultancy responds with frustration, the relationship degrades. If the consultancy complies silently, the risk shifts into delivery, and blame becomes likely.

‍

Mechanism-based response:

1. Make decision rights explicit. The client owns the decision, but the consultancy owns recommendation quality and risk transparency (Rogers & Blenko, 2006).
2. Require a decision packet for one-way-door choices. Options, risk, SLA impact, and rollback plan.
3. Capture risk acceptance. If the sponsor chooses the higher-risk option, document the acceptance with name and date.
4. Define escalation. If the decision breaks non-negotiables, escalate early, not after the incident.

‍

This approach respects client authority while protecting decision quality.

‍

Measurable Outcomes

‍

If visibility and restraint are working as a system, you should see measurable movement within a quarter.

1. Decision reversal rate decreases. Fewer “final decisions” get re-opened.
2. Decision cycle time improves for high-impact decisions because roles are clearer.
3. Operational stability improves. Use software delivery performance measures such as change lead time, deployment frequency, change fail rate, and recovery time (DORA, 2026).
4. Unplanned work ratio decreases. Incidents and urgent fixes stop consuming planned capacity.
5. Ownership becomes legible. In planning and incident reviews, people can answer quickly, “Who owns this outcome,” and “What guardrail did we use.”

‍

A quick diagnostic you can run in a leadership team. Ask three leaders and three team members the same two questions.

1. What decisions do teams own end to end?
2. When do we escalate, and to whom?

‍

If the answers vary widely, you do not have a visibility problem. You have a decision-rights problem.

‍

References

• Bezos, J. (2016). 2016 letter to shareholders. About Amazon. https://www.aboutamazon.com/news/company-news/2016-letter-to-shareholders/
• DORA. (2026). DORA’s software delivery performance metrics. https://dora.dev/guides/dora-metrics/
• Edmondson, A. (1999). Psychological safety and learning behavior in work teams. Administrative Science Quarterly, 44(2), 350–383. https://doi.org/10.2307/2666999
• Greenleaf, R. K. (1970). The servant as leader. The Greenleaf Center for Servant-Leadership. https://greenleaf.org/what-is-servant-leadership/
• Marquet, L. D. (2013). Turn the ship around! A true story of turning followers into leaders. Portfolio.
• Perna, G. (2019, November 4). Mission command requires sharp commander’s intent. The United States Army. https://www.army.mil/article/215297/mission_command_requires_sharp_commanders_intent
• Rogers, P., & Blenko, M. (2006). Who has the D? How clear decision roles enhance organizational performance. Harvard Business Review, 84(1), 52–61. https://pubmed.ncbi.nlm.nih.gov/16447369/

‍