THE IDEA
Act first, understand second
In a complex system, you can’t analyse your way to the right answer. The system has too many interacting variables, too many adaptive agents, too many emergent possibilities. The expert analysis that works beautifully for complicated problems produces false confidence when applied to complex ones. You need a different approach.
Probe-sense-respond is that approach. Instead of analysing the system to determine the correct action, you act first - but small and safe. You introduce a probe: a modest intervention, a small experiment, a gentle perturbation. Then you sense what happens. How did the system respond? What shifted? What was surprising? What feedback did the probe generate? Then you respond to what you’ve learned, amplifying what worked and dampening what didn’t.
The order matters. In complicated domains, you sense first (gather data), analyse second (find the answer), then respond (implement). In complex domains, you probe first (act to learn), sense second (observe the response), then respond (adapt based on what you learned). The probe isn’t a guess. It’s a deliberate act of learning that generates information the system couldn’t have provided without being disturbed.
IN PRACTICE
Learning by doing, carefully
A community organisation wants to increase engagement among young people. The complicated approach: survey young people, analyse their needs, design a comprehensive programme, launch it. The probe-sense-respond approach: run three different small events in three weeks - a sports evening, a creative workshop, and a drop-in pizza session. See which ones young people actually show up to, what they say about them, what energy they bring. Sense what the community responded to. Then invest in the format that worked rather than the one the analysis predicted would work.
A company entering a new market. The complicated approach: commission market research, build a detailed entry strategy, invest heavily in the plan. The probe approach: launch a minimal product in one small segment, see how customers respond, learn what the research couldn’t tell you (because the customers are adaptive agents, not data points), and build the real strategy from what you learn. The probe costs less than the research and produces better information because it tests reality rather than opinions about reality.
A person trying to establish a new habit. The complicated approach: research the optimal routine, design the perfect schedule, commit to the full plan. The probe approach: try three different versions for a week each. Morning meditation, lunchtime walk, evening journaling. See which one the real system (your life, your energy, your family’s schedule) actually accommodates. Then commit to the one that survived contact with reality.
WORKING WITH THIS
Designing good probes
A good probe is small enough to be safe, specific enough to be informative, and designed to generate feedback. Before probing, decide: what will we look for? How will we know if this is working? How quickly will we see a response?
Run multiple probes in parallel when possible. In a complex system, you don’t know which approach will work, so testing several simultaneously increases your odds and accelerates your learning. Each probe should test a different hypothesis about how the system might respond.
The hardest part is cultural. Probe-sense-respond requires comfort with not knowing the answer in advance. It requires leaders who can say “we’re trying this to learn” rather than “this is the plan.” It requires organisations that fund experiments, not just implementations. The payoff is better outcomes - because the strategy that emerges from probing is adapted to reality, not to a prediction of reality that was probably wrong.
THE INSIGHT
The line to remember
In a complex system, the smartest thing you can do isn’t think harder. It’s try something small, watch what happens, and let the system teach you what works.
RECOGNITION
When this is in play
You need probe-sense-respond when analysis keeps producing plans that fail on contact with reality. When experts disagree about what will work. When the system is too adaptive and interconnected for any model to predict its behaviour. When the cost of a small experiment is far less than the cost of a large plan that might be wrong. When someone says “we need more data before we act” about a situation where the only way to get the right data is to act.