The change keeps sliding back

The first force you'll meet. Large systems resist change - not because people in them are resistant, but because the structures, processes, habits, and incentives have mass. Moving them takes sustained effort, and the system is constantly pulling back toward what it knows. Understanding inertia doesn't make it disappear, but it changes your expectations about how much force is needed and for how long.

Why the pull-back happens in the direction it does. Where a system can go depends on where it's been. Past decisions, investments, and habits create grooves that are easier to follow than to leave. The system isn't sliding back to a random place - it's sliding back along the path it already knows.

An unexpected antagonist. Resilience is usually a good thing - the ability to absorb disturbance and return to function. But if the system you're trying to change is resilient, that resilience works against your intervention. The system absorbs the change and bounces back. The very thing that makes it robust makes it hard to shift.

What you need instead of brute force. Rather than trying to overpower the system's resistance, build the system's own capacity to adjust. Adaptive capacity is the resource that lets a system respond to new conditions - including the new conditions your change is trying to create. Build it, and the system moves with you rather than against you.

The shift from "are we doing this right?" to "are we doing the right thing?" Single-loop learning adjusts methods. Double-loop learning questions assumptions. Change that only operates at the first loop will eventually be pulled back by the unchanged assumptions underneath. Lasting change requires the second loop.

Small structural changes that make the desired behaviour the easy behaviour. Rather than constantly pushing for change through motivation and effort, a catalytic mechanism redesigns the conditions so the change sustains itself. The best ones are so simple they feel obvious in hindsight.

When you don't know what will stick, run several small experiments rather than one big bet. Each probe tests how the system responds. Some will fail - that's built into the design. The ones that find traction can be amplified. This approach works with the system's unpredictability rather than pretending it doesn't exist.

The smallest change that could shift the system. Not because bigger interventions are wrong, but because complex systems respond unpredictably to big pushes. Starting small lets you see how the system responds, learn from its reaction, and adjust before committing more. The change that sticks is often the one that was designed to be just enough.