Every other planning tool lets you model forward. Goal Seek runs backward: a CEO states "$2.4M EBITDA in twelve months," and the system returns three to five lever combinations that reach it — ranked by feasibility against the operator's own history.
The hard problem isn't running a projection. It's knowing which projections are realistic. We've solved that by calibrating every proposed lever value against the operator's own twenty-four-month history.
Pick a metric (EBITDA, margin, cash, revenue), a value, and a horizon. Optionally constrain which levers are allowed to move ("no headcount cuts") and set per-lever bounds. Or just type the prompt: "get me to $2M EBITDA without cutting headcount."
The AI generates eight to twelve candidate lever-value arrays, the simulator validates each against your real books, and the ranker keeps the top three to five distinct solutions. End-to-end: sixty to one hundred twenty seconds.
A lever value at the historical mean scores 1.0. A value at ±1σ scores ~0.7. A value at the historical max scores ~0.4. Outside the envelope — we drop the candidate. Least operationally painful path first.
Other products will use the same model. We will use the same model. The differentiation is not the LLM — it's the closed-loop feedback that calibrates every proposed lever value against the operator's own driver envelopes.
Every operational driver — enrollments, headcount, average ticket, occupancy, churn — carries a historical envelope: min, max, mean, volatility, computed nightly from your trailing twenty-four months. Every lever the AI proposes binds to one of those drivers. Every proposed value falls inside the envelope, or it gets dropped.
When the operator runs a scenario and time passes, the forecast-accuracy tracker compares projected to actual. Drivers whose AI projections were persistently wrong get larger uncertainty in their next envelope. The next path the AI proposes is better-calibrated than the last.
The AI is required to say "no realistic combination reaches this target" when that's true. Goal Seek won't manufacture a feel-good answer to a question your business can't actually answer.
When a run is marked infeasible, the page shows the closest path the simulator could find, the specific lever that broke realism, and the historical percentile at which that lever would need to move. It's the most useful answer possible: "here's what you'd need to do, and here's why we don't believe you'd actually do it."
A Goal Seek solution becomes a named scenario — with all the lever values frozen, the EBITDA bridge embedded, and the full P&L / BS / CF projected at entity and consolidated levels. When time passes, the forecast accuracy tracker measures how well your chosen path held up against the actuals.
Then those measurements feed back into envelope recalibration. This is what "the more the platform sees, the more credible its proposed paths become" means in practice.
Connect your books in an afternoon. See the diagnostics and the first scenarios on your real data, the same week. No implementation fees, no six-month rollout, no SOW.