Every plan of consequence in your company is a single number. Revenue will land at 104, the transformation will cost 12.5, go-live is the first of October. Everyone experienced in the room believes two things at once. The number is almost certainly wrong, and the year will be managed against it anyway.
That is not a failure of skill; it is a failure of form. The future has many ways of arriving, and a form that permits exactly one number forces everyone to pretend otherwise, then spend the year explaining variances against a fiction. The honest form of a plan is a distribution. It shows the range of outcomes, each carrying odds, updated as evidence arrives.
Odds, not lists
Most planning teams believe they already handle uncertainty, because they produce scenarios. Base, upside and downside make three spreadsheets and three stories. But when the downside case carries no odds, nobody can say whether it deserves a contingency budget or a shrug, so the decision defaults to temperament. Whoever is most senior settles the argument, and evidence never enters it.
Attach odds and the conversation changes. A 12% chance of missing the covenant is an instruction. A "downside scenario" is a mood.
The meeting, restructured
The deepest change is in what the monthly meeting argues about. Against a point plan, the meeting litigates whose number is right, which cannot be resolved and therefore never ends. Against a distribution, nobody's number is right by construction, so the argument moves to thresholds and interventions, where it can be won. At what probability of missing the year do we act? Which move shifts the odds most per unit of cost? Those questions have answers, and the answers can be scored afterwards.
This is also the only structure in which success stops being a surprise. Odds that update daily cannot ambush you in November, because the drift was visible in March.
Why the market stopped short
"Probabilistic" is everywhere in enterprise software now, so look at where the word points. It almost always describes model behaviour, outputs to be bounded and governed. Anaplan's agentic enterprise positioning pairs deterministic calculation with probabilistic AI as a trust story. Nobody publishes a distribution over plan outcomes; nobody states the odds that the year lands, updates them, and scores them against what happened.
The absence is not technical. Stated odds can later be checked for calibration, and that check is public and unforgiving. A vendor who says "probabilistic" commits to nothing. A vendor who states the odds can be proven wrong.
Horizon is built on the second posture. Plans carry distributions, commitments carry rolling odds updated daily, and the odds are scored against outcomes, so their calibration is itself on the record. That is the moat; a marketing team cannot retrofit checkable odds. To see your year as a distribution, start here.