When managing or intervening in complex systems (like human teams, software architectures, or biological organisms), our linear instinct is to control and prescribe. We attempt to dictate every step. However, because complex systems are adaptive and self-organizing, direct control almost always fails, producing unintended side effects.
Instead of prescribing solutions, a systemic approach uses constraints as tools.
A constraint is a boundary, a limit, or a rule of engagement. It does not dictate what must be done; instead, it narrows down the space of what can be done. By removing default, low-effort strategies, constraints force agents to explore and self-organize around new pathways.
This principle is universally applicable across domains:
- In physical training: If an athlete always performs a movement with compensation (e.g., relying on their lower back rather than their hips), a coach can introduce a constraint—like placing a physical block behind their heels or fixing a posture. The default strategy is blocked, forcing the nervous system to explore and discover new muscle coordination.
- In software teams: Limiting Work In Progress (WIP) in Kanban or scheduling a strict 15-minute time limit for meetings (constraints) forces the team to prioritize and coordinate differently, leading to spontaneous process improvements.
- In AI prompting: Designing a system prompt is not about listing every single action an agent must take. It is about establishing boundaries, feedback loops, and focal points (constraints) that allow highly adaptive behavior to emerge safely.
Using constraints is the difference between forcing a system to behave and creating the conditions under which the system naturally self-organizes in the desired direction. It is the practical execution of influence over control.