Openness & Psychological Safety

What this dimension measures

This dimension looks at the path unwelcome information takes through an organization and at what it costs the messenger. When an initiative is not working, does that news reach the sponsor, or does it stop at a manager who smooths it over? When it does reach the sponsor, does the plan change, or is the information acknowledged and ignored? And what happens afterward to the person who brought it: are they treated as helpful, or do they learn to keep quiet next time?

The assessment asks respondents to recall the last time someone brought unwelcome news about an initiative to a senior sponsor, and what happened both to the initiative and to the person. That pairing is deliberate. An organization where bad news reaches the top but changes nothing has one kind of problem; one where it never reaches the top has another; one where the messenger pays for it has a third.

An organization with high Openness tends to have problems named early, objections that actually alter plans, and people who can say "I do not know" or "this is not working" without consequence. An organization sitting lower on this dimension may look calmer from above, because its difficulties are real but invisible.

This is about the treatment of the messenger, not the treatment of the trial. Whether a failed test is stopped and learned from is a matter for Experimentation & Learning.

Why this matters for AI

AI will make mistakes, and some of them will be plausible, confident, and hard to spot. It will produce unexpected results. It will work well in a demonstration and poorly on the real task. The most reliable defense is accurate feedback from the people closest to the work, delivered early and acted on. That is exactly what this dimension measures.

Consider what happens without it. A team rolls out an AI tool. The people using it notice that it gets a certain class of case wrong. In an open organization, that observation travels upward within days, the tool's use is bounded, and the error becomes a lesson. In a closed one, the people using the tool work around it quietly, the sponsor hears that adoption is going well, and the error is discovered later by a customer, an auditor, or a consequence.

Success theater is the specific hazard. AI projects attract attention, and attention creates pressure to report progress. Where candor is unsafe, every effort looks like it is working until it visibly is not, and leaders hold a picture of adoption far rosier than the people doing the work. Decisions get made on that picture.

Openness also governs whether an organization can admit uncertainty, which AI demands constantly. Nobody knows in advance exactly where a tool will help. An organization where "we are not sure yet" is a safe thing to say can proceed by learning. One where it is not will either overclaim or avoid committing at all.

The leadership perception gap, the difference between what leaders believe about adoption and what their people see, is one of the most consequential findings an assessment like this can produce. It is largely a product of this dimension.

How this can work in your favor

When your responses describe people telling sponsors that an effort is not working and being treated as helpful for it, AI adoption gains an early-warning system.

Mistakes become lessons rather than costs. An AI tool's errors are reported by the people who see them first, while the error is still cheap. The tool's use is adjusted and the organization learns something about where it can be trusted.

Leaders see the real picture. Reports of progress reflect what is happening rather than what is comfortable to say. Decisions about where to invest next rest on accurate information, and the leadership perception gap stays small.

Uncertainty can be admitted. Teams can say they do not yet know whether a tool is worth it, which means they can run honest trials instead of defending premature commitments.

Assumptions get challenged in time. When an AI plan rests on a belief about the work that the people doing the work know to be wrong, someone says so before the plan is built on it.

Disappointing results survive. A trial that did not work is reported as such, which is more valuable than a trial quietly reframed as a success. The organization can stop, revise, and redirect effort instead of carrying dead projects.

How this can make AI adoption harder

When your responses suggest that unwelcome news does not reach sponsors, or reaches them without changing anything, or costs the person who carries it, AI adoption tends to fail in ways that stay invisible until late.

Problems stay hidden. The people using an AI tool notice its errors and work around them rather than report them. The tool keeps being used in ways it was never meant for, and the first clear signal is a consequence downstream.

Success theater. Every AI effort appears to be going well, because reporting otherwise is costly. Leaders hold a far rosier picture of adoption than the people doing the work, and the gap widens with every status update.

Commitment outruns evidence. Because disappointing results are softened or withheld, the organization keeps investing in efforts that the people closest to them know are not working. The eventual correction is larger and costlier than it needed to be.

Assumptions go unchallenged. AI plans are built on beliefs about the work that nobody dares contradict. The plan fails on contact with reality, and the failure is blamed on the tool.

Uncertainty is unsayable. Since "we are not sure" is unsafe, teams either overclaim what a tool can do or avoid trying it at all. Honest trials become impossible.

This friction is especially hard to see from above, which is part of what makes it consequential. The organization may feel calm precisely because the difficulty is not being surfaced.

What this dimension is not

Openness & Psychological Safety is not constant disagreement or a culture of complaint. The question is whether difficult truths reach the people who need them and change what happens next, not whether people argue.

It is not the same as running experiments. An organization can conduct many trials and still hide the disappointing results; openness is about what happens to the result, not whether the trial ran.

It is not a measure of niceness. A pleasant, conflict-free organization can sit low on this dimension if the pleasantness is maintained by keeping problems quiet.

It is also not a judgment of individuals. The dimension describes what the environment makes safe to say, as the respondent experiences it, rather than how brave any one person is.

What we look for

The assessment looks for the actual journey of unwelcome information rather than statements about open communication. Signals that tend to be informative include:

• The last time someone told a senior sponsor that an initiative was not working: did the news arrive, did the plan change, and what happened to the messenger?
• Whether objections raised by the people doing the work have altered a plan in recent months.
• Whether a trial has been reported as a disappointment, or whether every trial has been reported as some kind of success.
• Whether people can say "I do not know" in front of leadership without it costing them.
• Whether there is a known difference between what leaders believe about an effort and what the people running it believe.
• Whether someone who raised a problem last year is still raising problems this year.

"We have an open-door policy" is not evidence. What walked through the door, and what happened to it, is.