Built to be checked.
Every structure the platform runs, every engine that computes it, and where the AI layer stops — written down, so the method can be examined.
200+ calculated metrics — traceable.
Multiple deterministic engines compute the figures. The language model does not do the arithmetic and cannot change a result.
The same inputs return the same result, and any value can be redone by hand from the published formula.
The platform computes, interprets and writes. It does not decide, and the line between the two is documented.
You approve. The engines compute. You decide.
Four steps, always in the same order. The first and the last are yours.
Nothing enters without your check. Figures that do not reconcile are flagged before you apply — you decide, and nothing is altered.
The same input returns the same result, every time. The language model does not do the arithmetic.
It interprets the result, compares scenarios and writes the document. It changes no figure.
The platform delivers the material for the decision. The decision stays yours.
Every statement is checked against itself.
Before anything is applied, deterministic checks test each statement for internal consistency — the same way, every time. What reconciles passes; what does not is shown to you, with the figures involved, and you decide. Nothing is corrected on your behalf.
It reads, explains and writes. It does not invent numbers.
The concern is reasonable and the product was designed around it. The division is this:
- Reads and structures your documents
- Explains what the figure means for this company
- Writes the memo, the deck and the report
- Finds the clause in the room and names the file
- Every ratio, valuation and projection
- Every scenario and sensitivity
- Every figure that ends up in a document you sign
Redo any figure by hand.
The standard we hold to: a controller takes any value on any screen and rebuilds it from scratch. Pick the figure, open its calculation reference — inputs, formula and assumptions on one page, each input traced to a recorded statement — and redo the arithmetic. It resolves to the same figure.
Four documents, four questions.
Architecture, configuration and the deterministic engine — where the formula ends and the AI layer begins.
Product scope, calculation methodology, and how our claims can be verified.
Formulas as implemented, data lineage and controls — for technical review and audit.
The third parties involved in processing data and what each one does.
Talk to the team
Questions on method, scope or technical evaluation — talk directly to the people who built it.
Talk to the teamSecurity and controls
Isolation, encryption, permissions and the audit trail — the full set is on Trust.
See Trust