top of page

The Gap Between detecting a problem and deciding what to do about it.


Last week, the argument was that governance fails not because people lack training or awareness, but because organisations are not structurally designed to convert signals into decisions.


This week, I want to go even deeper into the question.

Why does escalation break even when people find the courage to speak up?



The credible intervention gap


Here is the uncomfortable finding that sits underneath most escalation research.

In regulated organisations, the top barrier to escalation is often not just silence. It is a rational calculation.


People weigh the personal cost of speaking up - reputational risk, career impact, being labelled difficult or commercially naive - against their estimate of what speaking up will actually achieve. If that estimate is low, silence becomes the rational choice.


Research on employee voice consistently identifies what is called learned futility as one of the most corrosive blockers. People do not repeat escalations that led nowhere. When prior experience showed that raising concerns produces documentation rather than decisions, the next concern stays local.


MIT Sloan research involving over 1,000 executives from 500 companies found that 84% regularly dealt with conflicts requiring escalation, yet relatively few organisations had effective structures to ensure issues reached the right decision-making level. When escalation architecture was missing, the pattern was consistent: poor decisions from incomplete information, damaged trust between colleagues, and compromised execution.


This means most regulated organisations do not primarily have an escalation problem. They have a credible intervention problem.


People stay silent less because they do not know the rule and more because they do not believe the system can absorb their concern without punishing them or wasting their effort.


That distinction matters enormously for how organisations respond. Training people to speak up is not the answer to a credible intervention deficit. Designing the system so that speaking up reliably leads somewhere is.



Seven structural blockers worth naming


Research across organisational psychology, regulatory supervision, and governance failures identifies the blockers that recur most consistently. None of them is primarily about individual awareness or courage.


Fear of personal cost. The calculus is usually not "I don't know the rule" but "raising this feels risky." Research consistently shows employees hesitate to speak up due to fear of damaging social relationships or being perceived negatively, especially when addressing issues with supervisors who control resources and career opportunities (Morrison, 2014; Detert & Burris, 2007).


Hierarchy and power distance. Early signals usually appear at the operational edge of organisations. Escalation requires challenging someone more senior. Without structural protection, signals stall in middle layers rather than moving upward.


Low psychological safety in everyday interactions. This is distinct from formal speak-up channels. When ordinary questioning is punished or subtly penalised, people reserve escalation only for extreme cases. ECB supervisory guidance explicitly links psychological safety, incentives, and speak-up culture in regulated firms.


Ambiguity about thresholds, routes, and ownership. OECD integrity research shows that reporting drops significantly when people are uncertain about what qualifies, where a concern should go, or who owns the response. Many organisations believe they have solved escalation by creating a hotline. They have not.


Learned futility. Voice research shows that people speak up more when they believe their concern will be taken seriously and acted on. It drops sharply when prior experience taught them that escalation disappears into the process. Feedback matters - when people raise concerns and never hear what happened, silence grows.


Incentives that punish friction. Many regulated organisations tell people to escalate while rewarding speed, delivery, and smooth execution. ECB guidance on governance and risk culture explicitly connects incentive structures, challenge, and speak-up conditions.


Weak managerial response capability. The immediate line manager matters more than most governance frameworks acknowledge. Research consistently shows that the local climate determines whether escalation feels safe or futile - often more than enterprise values or board commitments.



The Boeing pattern


The Boeing 737 MAX investigations found that internal engineers had raised concerns about the flight control system years before the accidents. The signals existed. Concerns circulated inside the organisation.


What failed was the structure that should have ensured those concerns reached an authority capable of intervention.


The organisation had oversight processes. It did not have a reliable escalation architecture.


A 2025 scoping review published in Business and Information Systems Engineering, analysing 141 studies on AI failure, found the same pattern across sectors: escalation and accountability gaps appear as root causes across technical, interactional and ethical failure categories. The failures are structural and predictable.



Three metrics that almost no organisation measures

Most organisations measure incident counts and performance outcomes. They are not mature in measuring the three metrics that actually reveal escalation quality:


Time to escalation. The interval between the first detection of a concern and formal escalation. In most organisations, this number is unknown because it is unowned.


Time to intervention. The interval between escalation and decision or action. This measures the governance response speed, not the technical one.


Escalation survival rate. The proportion of concerns that reach decision-making authority without being absorbed, downgraded, or lost in the process. This is the metric that most directly reveals whether the escalation architecture is functional or symbolic.


Deloitte's 2025 Global Boardroom survey found that 66% of boards report limited to no AI knowledge or experience, and nearly one in three say AI does not appear on their agendas at all. If AI risk is not on the agenda, escalation metrics almost certainly are not either.


These three numbers are not technical metrics. They are governance integrity metrics. They reveal whether authority is real or symbolic.



A four-part test

For escalation to work under live pressure, four conditions must exist simultaneously:


  1. Signal detection. Someone recognises when something looks wrong.

  2. Voice safety. They feel able to raise it without personal cost.

  3. Escalation pathway. The concern can travel clearly through the organisation to the authority.

  4. Intervention authority. Someone with the power to act receives the signal and can act.


Most governance investment goes into the second condition, while the failures occur in the third and fourth.

The question organisations ask after something goes wrong is usually: why did people not speak up?


The harder and the more revealing question is this: If someone had spoken up, would

the organisation have been able to act?


Escalation does not begin with courage alone. It begins with design.


So ask yourself:


Which of these four conditions is weakest in your organisation right now?


Is that weakness visible to the people responsible for governance - or only to those living with the consequences?


And most critically: are you measuring the three metrics that reveal whether your escalation architecture is real or symbolic?


Part of a 12-week series on AI governance and organisational readiness.

Dr Joanna Michalska is the founder of Ethica Group Ltd, which advises boards and C-suite leaders on decision authority and governance architecture under automation.


Sources:


Business and Information Systems Engineering, Springer 2025 - scoping review of 141 AI failure studieshttps://link.springer.com/article/10.1007/s12599-025-00970-2


Deloitte Global Boardroom AI Survey


Hughes, J. & Weiss, G. (2025). Formalise Escalation Procedures to Improve Decision-Making. MIT Sloan Management Reviewhttps://sloanreview.mit.edu/article/formalize-escalation-procedures-to-improve-decision-making/


Morrison, E.W. (2014). Employee Voice and Silence. Annual Review of Organisational Psychology and Organisational Behaviour, 1, 173-197https://www.annualreviews.org/content/journals/10.1146/annurev-orgpsych-120920-054654


Detert, J.R. & Burris, E.R. (2007). Leadership Behaviour and Employee Voice: Is the Door Really Open? Academy of Management Journal, 50(4), 869-884https://journals.aom.org/doi/10.5465/amj.2007.26279183


OECD integrity and reporting researchhttps://oecd.ai/en/ai-principles

ECB supervisory guidance on governance and risk culturehttps://www.bankingsupervision.europa.eu/


Amy Edmondson, psychological safety and voice research, Harvard Business Schoolhttps://www.hbs.edu/faculty/Pages/profile.aspx?facId=6451



 
 
 

Comments

bottom of page