How it works
Each engine runs established mathematical methods on the organization’s own numbers. The AI works around the engine: it researches, challenges and explains, but it never changes a number and never makes the decision.
The engines
An AI assistant can describe these methods. The engines run them, on the decision in front of the team, with its own competitors and its own numbers.
Discrete-choice modelling, the method for which Daniel McFadden received the Nobel Prize, estimates how customers choose between competing offers on price, value and position. It gives the chance of winning or keeping a customer at each price.
Used in pricing and sales bids
Game theory models what a competitor will bid given what it knows, what a supplier will concede given its alternatives, and what a buyer will accept. The engines find the point at which neither side gains by moving.
Used in sales bids, procurement and strategic commitments
Decision analysis lays out the choices and the uncertain events in the order they occur. Simulation runs thousands of scenarios across the ranges the team sets, so each path is valued across all of them rather than on a single forecast.
Used in strategic commitments, market entry, bids and procurement
The AI
From a plain-English description of the decision, it drafts the options, criteria and scenarios for the team to confirm.
It gathers the facts and evidence the brief is missing.
Before anything is committed, it tests the assumptions, argues the other side’s position and asks which inputs are driving the result.
Once the engine has run, it gives an independent view (concur, concur with conditions, or dissent), ranks the risks, and lists what the decision owner must still decide.
It drafts the brief for leadership or the board from the same record, in the version each audience needs.
Every decision and every override is recorded with its reason, so the next decision starts from what was learned on the last one.
The engine is the authority on the numbers. The executive is the authority on the decision.
The difference
| Who produces the numbers | Can the result be checked? | Who decides | |
|---|---|---|---|
| Prompting a model | The model, and they can change with the wording of the question | Hard to reproduce | Whoever accepts the answer |
| An AI agent given an objective | The agent | Through its logs | The agent, within the limits it was given |
| AI inside a pricing or procurement platform | The platform’s models, trained on transaction history | Within the platform | Often automated |
| Decisiums | A purpose-built engine, on inputs the team has confirmed | Every input is listed and sourced; the same inputs always give the same result | The executive |