Scenario planning
Not "what will happen" — "what could happen, and is our plan still standing if it does?"
What scenario planning actually is
Scenario planning isn't forecasting. Forecasting tries to predict the single most likely future; scenario planning deliberately builds several different, internally consistent futures and asks whether today's plan survives each one. The output isn't a prediction — it's a stress test.
That distinction matters because the future that actually shows up is rarely the "most likely" one anyone predicted. Planning against a range means fewer total surprises, even if no single scenario turns out exactly right.
How AI generates alternative scenarios
Left alone, most people default to one or two futures — usually "things continue as they are" and "the thing I'm already worried about." AI is useful here specifically because it isn't anchored to one narrative: it can generate a wider spread of plausible futures quickly, including the ones nobody in the room would have raised first.
A worked example: the 2x2 scenario matrix
Pick two genuinely uncertain forces that matter most to a decision, cross them as two axes, and four distinct futures fall out automatically — each internally consistent, each worth planning against.
Two axes — regulatory pace and AI capability growth — produce four distinct worlds: Cautious Adoption, Rapid Transformation, Status Quo, and Disrupted Incumbents. A plan that only survives one of the four is a bet, not a strategy.
Comparing outcomes across scenarios
This is a different comparison than ranking decision options. Here, the decision usually stays fixed while the external world varies — the question is whether the same plan holds up in "Rapid Transformation" as well as it does in "Status Quo." AI can run the same plan mentally against each quadrant and flag exactly where it would break.
Handling uncertainty honestly
Resist the urge to assign a false-precision percentage to each scenario — "35% likely" sounds rigorous but is usually a guess wearing a decimal point. Talk in terms of plausibility instead: which futures are genuinely conceivable given what's known today, not which one you'd bet on if forced to pick just one.
What "the best scenario" actually means
The goal usually isn't picking the single scenario to plan for — it's finding the strategy that performs acceptably across most of them. A plan that wins big in one quadrant and fails badly in the other three is riskier than a plan that does reasonably well in all four, even if the first one has a higher ceiling.
Where AI's sense of the future breaks down
Anchored to the past
- Every pattern it knows comes from what already happened
Blind to structural breaks
- Sudden regime changes don't look like anything in the training data
Weak on non-linear change
- Compounding, tipping-point dynamics are notoriously hard to reason about well
How this should actually shape a decision
Scenario planning isn't meant to tell you which future will happen — it's meant to reveal which parts of today's plan are fragile, and which are genuinely robust regardless of how things unfold. Use it to identify the one or two assumptions your whole strategy is quietly resting on, then decide if you're comfortable with that exposure.
The short version
Scenario planning trades the comfort of one confident prediction for a plan that survives being wrong about the future — which, statistically, everyone eventually is. AI is genuinely good at generating a wider spread of plausible futures than one room full of people would think of alone, and at checking a plan against each one quickly. It's weak exactly where the future stops resembling the past. Use the scenarios to pressure-test a plan, not to pick a single future to bet everything on. For ranking risks within one specific decision rather than mapping external futures, see Risk Analysis.