Complex Adaptive Systems
Complex Adaptive Systems thinking treats an organisation as agents interacting under local rules rather than a machine you can redesign from the top, and the practical move it recommends is small safe-to-fail experiments instead of one big plan.
A loop keeps turning, a small probe tried, its effect sensed, then amplified or left to fade.
Reach for this when…
- A carefully detailed rollout plan keeps failing in a different way at every site.
- Leadership keeps trying to mandate a behaviour that only spreads when people choose it themselves.
- A problem changes shape every time someone tries to fix it directly.
How to run it
- Map the system: who are the agents, what rules do they actually follow, what patterns are already emerging.
- Check whether the problem is genuinely complex (Cynefin) rather than merely complicated - a complicated problem still has a right answer to find.
- Run several small, cheap, safe-to-fail probes rather than one large planned intervention.
- Sense what actually happened in each probe, honestly, including the ones that did nothing.
- Amplify what's working and dampen what isn't, then probe again.
A worked example
Situation. Giulia Rossi runs operations for Alta Collina, a small wine growers' cooperative near Bologna, Italy, and had watched three separate five-year irrigation plans fail in three different unpredictable ways across the co-op's member farms.
Applied. She stopped planning the rollout and ran three small pilots on different farms instead, then watched rather than pushed - one pilot started spreading between neighbouring farms on its own, by word of mouth, and she quietly let the other two drop.
Result. Adoption across the co-op ended up higher a year later than any of the failed master plans had achieved, because it grew from a pattern that was already working rather than one imposed from above.
The catch
This lens is easy to use as an excuse for not deciding anything - 'it's complex' can become cover for avoiding a call that actually was simple. It also takes longer to show results than a plan does, which is a hard sell to a board that wants a timeline.
If you can already see the right answer, this isn't a complex problem - don't dress up indecision as emergence.
Origin: Stacey; Snowden (Cynefin)