Psychological Safety in Executive Teams: Amy Edmondson

Psychological safety in executive teams concerns whether members expect they can raise work-relevant questions, mistakes, uncertainty and disagreement without humiliation or punishment for speaking up. Amy Edmondson’s research helps leaders examine those interpersonal conditions. The practical test is what happens when a contribution challenges a preferred plan.

Coachful is our #1 recommendation for organizing the executive’s preparation, chosen leadership actions and private coaching review. Use that workflow to examine actual responses and follow-through. It cannot certify a team as safe or replace the operational decisions, fair treatment and formal processes that the work requires.

Understand the concept and the original research

Edmondson’s 1999 study of psychological safety and team learning examined fifty-one teams in one manufacturing company. Psychological safety was associated with learning behavior; the model linked learning with performance. Team efficacy did not add the same association with learning when psychological safety was controlled.

The study combined methods, but its cross-sectional survey did not demonstrate causality. The paper also identifies limits concerning construct differentiation and generalization from one company. It is important foundational research, not a randomized test showing that an executive-coaching worksheet causes a particular performance gain.

In a 2022 Harvard Business School interview, Edmondson distinguishes psychological safety from accountability rather than presenting a trade-off between them. She describes framing work around learning and modeling curiosity. The original applications below develop those ideas through concrete leadership choices; they are not a validated team diagnostic or a prescribed intervention protocol.

A team-level concept also deserves more care than one leader’s impression. Members may experience the same meeting differently. A CEO’s willingness to disagree with peers says little on its own about another executive’s anticipated cost of challenging that CEO.

Keep candor, standards and authority together

A relevant concern can be heard without automatically determining the decision. Someone may identify a risk that further evidence does not support, or offer an option the team cannot use. A respectful response can explain the evidence, constraints and decision rather than treating rejection of an idea as rejection of the person.

Accountability needs an object: an agreed requirement, a decision, an information handover or a standard of conduct. “Be more accountable” gives little direction. “Send the pilot figures to the launch decision owner by Wednesday noon” can be checked and revised if the route fails.

Do not promise immunity from every consequence. Deliberate misconduct, a serious formal complaint and an uncertain operational experiment are different situations. Clarify which organizational route applies. A coach cannot grant employment protections or resolve a conduct issue merely by calling a conversation psychologically safe.

Useful leadership practice therefore asks two questions together: can relevant information be raised, and how will it be evaluated and acted on? An open conversation with no decision standard can drift. A demanding target with no usable way to report constraints can conceal information the target depends on.

Make interpersonal risk specific

Ask the executive to reconstruct a recent exchange. What was the decision? Who had relevant information? What did someone say, what response followed and what happened afterward? Keep remembered words, documentary facts and the client’s interpretations separate.

A statement such as “people are afraid to challenge me” needs exploration. The executive may have observed a late warning or heard direct feedback, but they cannot infer everyone’s private experience from silence. Someone may be uncertain, short of information, using another appropriate route or choosing not to speak in that setting.

Power can affect the practical choice. A leader might invite disagreement after stating that only a disloyal colleague would oppose the plan. The invitation exists, but so does the threatened social cost. Examine that sequence before asking others to become more courageous.

Offer appropriate participation routes. Written preparation, time to consider a question and a named follow-up conversation can support contribution. Do not make public disagreement the only acceptable form of engagement, or require personal vulnerability and failure stories as a test of belonging.

Frame an actual decision that can still change

State what is known, uncertain and open to influence. “We need to decide the first release size; support demand is uncertain” invites a more specific contribution than “Any thoughts?” If a decision is fixed, explain what remains open rather than staging consultation about a settled outcome.

Give members the material they need in time to use it. Ask which assumption could change the decision and who has relevant experience or evidence. A private concern about support capacity becomes more discussable when the team knows the forecast, available hours and decision owner.

Explain how input will be considered. A concern may lead to a smaller release, further analysis, a contingency or a reasoned decision to proceed. The team should know what evidence matters and when an answer will come back. An invitation without a response route can leave the work unresolved.

For meeting preparation, see the guide to creating a useful dialogue container. Use the practical process to hear the relevant contribution, while preserving the actual organizational authority for the decision.

Invite input and respond before defending

Use a question suited to the work: “Which assumption is least secure?” or “What would make this support plan fail?” Allow thinking time. Ask for clarification before judging whether the concern arrived late or whether its presenter should have known more.

Acknowledge the information without pretending it has already been verified. “That could change the plan; let us inspect the figures” is a substantive response. A quick thank-you followed by ignoring the issue gives the team a different lesson about the value of speaking.

If your first response discouraged contribution, repair it specifically. Name what you did, return to the question and complete the promised follow-up. An apology can matter, but the team also needs to see whether the actual decision process changes.

Keep disagreement reasoned. Ask what supports an alternative and explain why an option is accepted, declined or held. The aim is usable information and an accountable decision, not consensus at any cost or a rule that every objection must win.

Begin the fictional executive-coaching case

All people, data, meetings and outcomes in this case are fictional. Elin is CEO of a small software business. Her executive colleagues include support lead Nora, operations lead Tariq and finance lead Mateo. A first release to two hundred eligible customers is being considered.

Elin hires Mara for three private forty-five-minute coaching sessions on 9 November, 23 November and 7 December 2026. Mara coaches Elin’s leadership practice, not the absent colleagues. Elin shares only the permitted generalized account of the work. The company’s customer and operational records remain in its authorized systems.

The initial support forecast assumes twenty percent of the two hundred customers will need one support case: forty cases. Nora receives a pilot summary on 5 November showing twenty cases from fifty pilot customers, or forty percent. That small pilot may not predict the next group reliably, but it challenges the original assumption.

At the 6 November executive meeting, Nora raises it. Elin replies, “We agreed the date. Why are we hearing this now?” The conversation moves to the sales announcement without examining the figures. Elin later describes the meeting as open because Nora was allowed to speak.

In the 9 November coaching session, Mara asks what happened after Nora spoke. Elin recognizes the gap: the words were permitted, but the information did not enter the decision. Mara does not diagnose Nora’s feelings or declare the whole team psychologically unsafe from Elin’s account.

Mara: What answer did the team need before confirming the release size?

Elin: Whether support could handle the demand if the pilot rate continued.

Mara: What did your first response help establish?

Elin: Nothing about capacity. I challenged Nora’s timing before understanding the new information.

Elin chooses to reopen that specific decision and repair her response. She does not ask Mara to survey the team or classify each colleague’s willingness to speak. They agree to review the actual decision record and the feedback Elin can appropriately obtain.

Return to the concern and obtain the real constraints

At the 10 November executive meeting, Elin says: “My first reply challenged the timing before we examined your information. I want to correct that. Nora, please take us through the pilot figures. We will decide the release size after checking support capacity, and then clarify the reporting route.”

Nora confirms the twenty cases and fifty pilot customers. Mateo notes that this is a small pilot, not proof that the next group will have a forty-percent rate. They use it as a planning scenario alongside the original twenty-percent assumption, preserving uncertainty.

Tariq verifies that Nora’s support group has twenty protected staff hours for case handling in this release week after its other commitments, with reporting time allocated separately. No additional staff capacity has been approved. They use twenty-five minutes per ordinary case as a planning estimate, while recognizing that individual cases can take longer.

Forty planned cases require one thousand minutes, or sixteen hours forty minutes. The twenty-hour allocation contains twelve hundred minutes, leaving two hundred minutes of reserve. Eighty cases, the scenario for two hundred customers at the pilot rate, would require two thousand minutes and exceed that allocation by eight hundred minutes.

The group compares three options. Sending to all two hundred under the original estimate relies on the lower rate. Adding capacity cannot be treated as available because nobody has authorized it. Releasing to one hundred produces forty forecast cases under the higher scenario, leaving the same two hundred-minute reserve.

Nora says the smaller release is workable as a bounded plan, not a guarantee about demand. Elin, the authorized release decision owner, approves one hundred customers for 16–20 November. The other hundred invitations are deferred. The launch date alone does not decide how many people must enter.

Complete the decision record and reporting repair

The accepted record states: “Invite one hundred customers in the release week. Nora’s group has twenty protected hours, with forty ordinary cases planned at twenty-five minutes each and two hundred minutes of reserve. At thirty-five received cases, review remaining work and pause any further expansion pending a decision.”

Nora owns support-case totals and threshold reporting. Tariq checks available coverage and sends a coverage handover at 09:00 each release-week morning, including 17 November. Elin owns the expansion decision. They will use the company’s current operations channel, not private coaching notes. Nora accepts the case-reporting duty. Tariq explicitly accepts the timed coverage handover within his existing, separately allocated reporting time. Both confirm capacity, and Mateo confirms that the record preserves the forecasting uncertainty.

They separately inspect why the pilot summary did not reach the decision earlier. Elin finds that the meeting pack required a sales update but had no named route for pilot support results. Nora had the summary only the day before the meeting. Elin accepts responsibility for the pack design rather than treating the warning as proof of Nora’s poor commitment.

They add a support-results field with Nora as owner and an explicit review before invitations are finalized. This does not excuse future missed agreed updates; it creates a requirement that can be understood and checked. Elin sends the completed record to the same executive group and all four confirm it reflects the decision.

That acknowledgment closes the immediate decision loop. It does not establish that every member trusts Elin, that the new reporting habit will persist or that the entire business is psychologically safe. The team has completed one more useful response to a specific contribution.

Observe the release and retain a real lapse

During the release week, Tariq’s agreed coverage handover is due at 09:00 on 17 November. He misses it while handling another operational problem. Elin starts, “We cannot have this happening again,” then pauses and asks what information is missing and when the current status can be supplied.

Tariq acknowledges the missed coverage handover and provides the current coverage status at 09:20. The twenty-minute delay remains in the record. They agree that operations coordinator Luca will back up later coverage handovers. Luca actually accepts the role within the existing operations reporting allocation, separate from the twenty case-handling hours. The incident is neither erased nor turned into a judgment about Tariq’s character.

On 18 November, received support cases reach thirty-five. Nora uses the agreed route, and the executives review the work and available time. Elin confirms that no expansion to the deferred hundred will occur that week. The guardrail has led to an actual decision rather than a warning nobody owns.

By 20 November, forty-seven cases have arrived and been handled. Forty-six take twenty-five minutes each; one takes forty. The actual total is 1,150 + 40 = 1,190 minutes, or nineteen hours fifty minutes, within the twenty protected hours with ten minutes remaining.

The plan’s forty ordinary cases used one thousand forecast minutes. Actual work used another one hundred ninety of the two hundred reserve minutes. The larger case volume therefore came close to the boundary. It does not show the original forecast was accurate or that the team could absorb another hundred invitations.

The smaller release and available reserve help explain the work outcome. A different demand pattern could have exceeded capacity. The coaching conversation is not evidence that psychological safety alone caused the completed support work or removed technical and staffing risk.

Review different experiences instead of announcing transformation

At the 23 November coaching review, Elin brings the permitted generalized decision account. Nora has told her that inspecting the figures and returning a decision was useful. Nora also says she still hesitates to raise a concern immediately after Elin expresses a strong preference. Elin records both parts rather than converting thanks into a claim of complete trust.

Mateo values keeping the pilot uncertainty explicit and asks for the next release’s evidence before approving a forecast. Tariq acknowledges the missed coverage handover and says the accepted backup arrangement is more practical. None of these comments is treated as a validated psychological-safety score.

Elin chooses a revised opening for the next decision: “Here is the question and what remains open. I will hear the assumptions and constraints before stating my preference.” She sends that process proposal to the executives, who acknowledge it. The next release decision has not yet taken place.

By the 7 December closing session, the future decision remains pending while further support data are gathered. Elin keeps the revised preparation and response practice. Mara sends the agreed coaching summary through the individual private client conversation, and Elin confirms it reflects her choice.

The three sessions total one hundred thirty-five minutes. The service closes without claiming a lasting change in team climate or automatic extension for an unresolved business decision. Elin has a completed case to learn from, a retained lapse and continuing work whose effects still need observation.

Use a response record that supports learning

An original response record can make follow-through visible without assessing who is sufficiently candid. Keep it in the appropriate organizational system and adapt these fields to the work:

• Contribution: What relevant concern, question or correction was raised?
• Evidence and uncertainty: What supports it, and what remains untested?
• Immediate response: What did the leader actually do, including any repair?
• Decision route: Who can decide, what must be checked and when is an answer due?
• Returned answer: What was accepted, declined or held, and did the contributor receive it?
• Follow-through: What happened in practice, what failed and what remains open?

Elin’s record would retain the ignored initial concern, actual figures and capacity, accepted limited release, missed coverage handover, guardrail decision and mixed feedback. Removing the first mistake would make the story more flattering and less useful.

For a complementary leadership lens, see the leader-member exchange coaching guide. Keep the question distinct: dependable individual relationships can matter, but one strong relationship does not establish shared team expectations about speaking up.

Gather evidence without creating a courage ranking

Review actual exchanges and responses over time. More reported mistakes can mean more disclosure, more underlying errors or a change in how events are counted. Investigate the events and reporting process before announcing either improvement or deterioration from the number alone.

If using an established survey, check its suitability, terms, administration and interpretation with appropriate expertise. The original questions in this article are coaching prompts, not the published research scale. Do not reproduce an instrument or invent a threshold from a casual average.

Explain who will see responses and what decisions they support. In a small executive team, distinctive roles or comments can identify a person even without their name. Do not promise anonymity that the actual process cannot protect, and do not make a group exercise the only route for a concern about retaliation.

Use the relevant formal organizational process for alleged misconduct or retaliation. The leader can improve everyday inquiry and follow-through while preserving those routes. A reflection exercise cannot determine the facts of a complaint or supply protections beyond the practitioner’s authority.

Organize the executive’s practice in Coachful

Create an original private worksheet for the executive’s preparation: the decision, expected challenge, planned invitation and actual response. Deliberately deliver it to the intended client and confirm access. Keep company evidence and other people’s sensitive information in their appropriate systems.

Use a Task for a client-chosen action such as returning the decision explanation. A tick records the action; it does not establish that colleagues feel safe. Private journal entries remain distinct from a reflection the executive intentionally submits to the coach.

Coach review holds private notes for the authorized coach team, with Reviewed or Follow up. The client cannot see those notes and they are not included in the response PDF. Send agreed feedback separately, as Mara does at the close, and agree any sponsor summary and audience independently.

Check session recap arrangements before including sensitive team concerns. Recaps send automatically by default; approval settings and Session Type overrides affect review before sharing. Explore Coachful’s features around actual preparation and feedback needs rather than promising a native team-climate certification or employee monitoring tool.

Frequently asked questions

Must everyone agree before a decision is made?

No. Hear and evaluate relevant contributions, then explain the actual decision through the appropriate authority. Respectful disagreement can remain after a reasoned choice.

Does a friendly meeting prove psychological safety?

No. Examine what happens when someone risks disagreement or reports a mistake, including the later response and decision. Comfortable conversation alone does not establish those conditions.

Should leaders share personal failures to encourage others?

A relevant acknowledgment of a work mistake can help explain accountability. It should not become a demand that colleagues disclose personal experiences or prove vulnerability in public.

How soon should a team be declared safe?

Avoid that blanket declaration. Observe repeated responses, differences in experience and the limits of the evidence. One repaired meeting can be useful while substantial work remains.