We say we need more affordances and more flexible constraints, they think we need agency.

I don't know why it took me so long to realize that when the entire agility/coaching culture has devolved toward infantilization.

Instead, many people understand: "engineering teams need to be facilitated and guided in order to develop some degree of initiative so we don't need to babysit them anymore."

By team autonomy, we mean ensuring teams have all the capabilities they need and authority to take decisions regarding their value stream. This requires integrating business, product and design capabilities in engineering teams, tightly coupled to operations of their value stream.

Now that I started using them, I realized that spatializing and visualizing the decision process helps me both think and communicate.

As often, it turned out the most simple aspect, the visualization tools, were also the most powerful.

As I dug a bit further, started learning about complexity theory, the value of the underlying theories became apparent quite fast.

It forces us to ask the right questions, gives important heuristics, organizes our understanding and approach of the obvious in ways that force a paradigm shift.

More importantly, we must not consider deviation from the expected result as a risk. It can just as likely be an opportunity, resulting in innovation.

Fast-feedback to accelerate learning and mitigate the cost.

By adding affordances and enabling constraints, we change the system's dispositional state rather than trying to direct people's behavior in opposition to the forces at play.

Numerous and in parallel, because their scale allows us to explore multiple options at once, limiting the time we spend between stable dispositions of the system, which in turn limits the risk of falling back into our old ways.

Safe-to-fail means small experiments with a limited scope and impact. The ones that work can be fostered and scaled, the ones that fail can be stopped. In both cases, we learn more about the dispositions of our system, so we can conceive the next set of experiments accordingly.

How can we actually mitigate the risk?

We need to run numerous safe-to-fail experiments in parallel with short feedback cycles.

Ultimately, the tight constraints on the system erode and break, as they take too much energy to maintain, and the small improvements gained fade away.

In the meantime, the situation keeps deteriorating, and the issues of the existing situation take deeper root, with entropy causing sclerosis across the entire socio-technical system.

To mitigate the risk, we tighten control. Long, planned projects with tight constraints to direct people's behavior in order to limit unintended consequences. Deviation from the plan is seen as a threat. Handling changes one by one at the scale of the entire organization with a homogeneous approach.

This leads to a fear of failure that paralyzes initiatives, and leads to a "failure is not an option" mindset. Paradoxically, this approach to change management is more likely to fail.

Due to all these, implementing change is risky and often fails. And in many organizations, failures has an additional cost: reputation. Anyone conducting an initiative or a project that leads to failure will likely suffer repercussions. Sometimes a loss of influence, sometimes worse.

- Any action on a complex system leads to unforeseen effect, so the change can backfire and create a worse situation

- Socio-technical systems are resilient and resist attempts to pull them out of stable states.
- The systemic failure of large-scale, fast paced transformations made the industry wary of any attempt.