Every budget is a story about one future. The problem is that the future rarely cooperates.
Scenario planning is the discipline of telling several stories at once—and deciding in advance
what you’ll do in each. Once a luxury of large corporate FP&A teams, it’s now accessible to any
business, largely thanks to AI-powered planning tools.

The core practice is simple. Alongside your base plan, build at least two others: a downside
case (revenue 15–20% below plan, a key customer lost, costs inflating) and an upside case
(demand outruns capacity). For each, answer three questions. What early indicators would tell
us this scenario is unfolding? What actions would we take, and in what order? And where are
the trigger points—the specific numbers at which we act?

That last part is what separates scenario planning from worrying. “If monthly revenue drops
below $X for two consecutive months, we freeze hiring and cut discretionary marketing” is a
decision made calmly in October that would otherwise be made in panic in March.

Here’s where AI changes the game. Traditional scenario modeling meant maintaining three
fragile spreadsheets that drifted out of sync the moment anything changed. Modern planning
platforms hold one driver-based model and let you flex assumptions instantly: drag churn from
2% to 4% and watch cash runway recompute across all thirty-six months. What-if analysis that
took an analyst a day now takes a minute, which means you actually do it.

Machine learning adds two deeper capabilities. Probabilistic forecasting, including Monte
Carlo simulation, replaces single-point estimates with distributions—instead of “we project
$2.4M in cash by June,” you get “there’s a 90% chance cash stays above $1.1M,” which is a far
more useful statement for risk decisions. And sensitivity ranking identifies which assumptions

your outcomes are most exposed to. Businesses are often surprised: they obsess over revenue
while the model shows their survival is more sensitive to payment timing or gross margin drift.

Generative AI contributes at the edges—drafting scenario narratives, suggesting risk factors
common to your industry that you haven’t modeled, and translating model output into boardready language.

A warning label: models inherit the limits of their training data. A model trained on five calm
years will underestimate tail risk. Use AI to compute consequences; use human judgment to
imagine the scenarios worth computing.

You can’t predict the future. You can rehearse it. Businesses that rehearse respond in days
while competitors are still convening meetings.

Next week: the cash hiding inside your own balance sheet.