Strategy Engine

Decisions with the future attached.

Anyone can generate a plan. The Strategy Engine models what each plan would do to your company — then remembers what it predicted, and answers for it.

Problem

Confident documents, no future attached.

The market is drowning in confident documents. A plan without a predicted outcome is a mood board; a recommendation without recorded reasoning is a liability with good formatting.

Mechanism

Survey → predict → decide → compile.

The survey captures the goal (A to Z or Together — your call). Cortex asks the questions that change the decision. Then the engine proposes genuinely different paths and the Business World Model predicts each: ranges, assumptions, sensitivity. Approval writes the Decision Ledger. The blueprint compiles into missions.

Under the hood

The whole ladder, shown plainly.

Turns your live state into the representation the models reason over. Your company, as the model sees it.

Anatomy

The decision brief.

Every recommendation carries the same shape:

Decision brief

recommendation: one chosen path
alternatives: always ≥2 real ones
evidence: what supports it
assumptions & uncertainty: named, not hidden
expected outcome: a range, with a model version
cost / risk / reversibility: stated up front
what happens on approval: compiles into missions

Worked example

Aster’s three paths, live.

DemoLaunch budget

$8k$30k

Recommendation · $20k

Focused design-partner program

Alternatives: Services-led entry · Paid launch campaign
Evidence: 12 interview summaries + 3 comparable launches
Assumptions: Win rate holds near 30% (9 months closed data — confirm?)
Expected outcome: 8–14 qualified conversations, 30 days (range, model v1.4)
Cost / risk: Moderate cost · moderate risk · reversible within 2 weeks

Control

Predictions, preserved.

Predictions preserved verbatim at decision time · model versions visible · abstention is honest output, not failure · A-to-Z mode still stops at your decision gates.

UnderstandDecideDeliverVerifyLearn

Company State feeds the World Model; the Decision Ledger sharpens the next recommendation.

Hard question

How is this different from asking GPT for a business plan?

Three things a chat can’t have. First, your state: the engine reasons over your live, sourced company model, not a prompt’s summary of it. Second, a calibrated world model: predictions with ranges, versions, and a public benchmark — not fluent confidence. Third, consequences: the ledger remembers every prediction and checks it against reality, so the system gets sharper where a chat just gets newer. A plan you can hold accountable is a different product from a plan.