We knew before we started.

What we were told

A company asked us to help improve its software development organization.

Before the engagement began, we met with the CEO and the head of development to discuss where we could help.

We were also told that the access and permissions necessary to do the work would be granted immediately.

On paper, there was no access problem.

But during the same conversation, we observed something that made us doubt that prediction—not because we thought anyone was lying, but because their behavior pointed in a different direction.

What we observed

The CEO first suggested bringing us into an existing project that needed improvement.

The head of development explained why that would not work: the project already had a fixed release date.

The CEO suggested another project. This one had no fixed release date.

Again, the head of development explained why it would not work: too many stakeholders were involved for an outsider to step in and take responsibility.

Both objections could have been perfectly reasonable on their own.

So the CEO tried a third option: instead of assigning us a project, perhaps we could help improve the productivity of the development team.

This time, the response was positive.

That sounded promising.

So we asked a question.

Was there a shared development approach across the organization, or was each engineer free to decide how to work?

The answer was that engineers were intentionally given the freedom to choose their own methods.

We then asked the head of development a hypothetical question:

If someone could show you a different way of working that made you ten times more productive, would you adopt it?

The answer was no.

That answer mattered.

But the pattern mattered more.

The pattern

Every time the CEO proposed a way for us to contribute, the head of development explained why it could not work.

• Project A could not work.
• Project B could not work.
• Improving development practices initially sounded acceptable—but the person responsible for development then explicitly said that even a dramatically more productive way of working would not be adopted if it came from someone else.

There was another signal.

No alternative was proposed.

The conversation was ostensibly about finding a way for us to help. Yet one person was repeatedly explaining why each proposed path was impossible without suggesting a path that would be possible.

The individual objections were data points.

The pattern across them was evidence of something larger.

Our hypothesis

The problem was not Project A.

It was not Project B.

And it was not the particular form of productivity improvement being discussed.

Our hypothesis was that the development organization did not actually accept outside intervention.

If that hypothesis was correct, the problem would not end with deciding what work we should do.

Once the engagement began, the same resistance would appear in the things required to do that work—including access to systems, information, and permissions.

So before the engagement started, we predicted that obtaining the access necessary to work effectively would become a problem.

Why we proceeded anyway

Under normal circumstances, this would have been a reason to decline the engagement.

We intended to.

But business decisions are made by people, not optimization algorithms.

The engagement had come through someone to whom we owed a great deal. Given that relationship and the circumstances, declining was not realistically an option.

So we proceeded despite the risk we had identified.

What happened

After the engagement began, the access and permissions we had been told would be granted immediately became a significant obstacle.

The prediction was confirmed.

But this was not a lie-detection exercise.

We did not need to determine whether anyone had knowingly made a false statement. They may have genuinely believed that the necessary access could be provided immediately.

For our purpose, intent was irrelevant.

The statement was one data point.

The behavior we observed during the same conversation was another.

When those data points pointed in different directions, we formed a hypothesis about what would actually happen—and waited for observable evidence to confirm or reject it.

What people said was one data point. How they behaved while saying it was another.