Skip to main content
Agentica

Answer — AI Governance

What is Scoreable Proof?

Scoreable Proof is a governance history organized so any framework or algorithm — Law 25, the AMF guideline, OSFI E-23, or a relying party’s own model — can compute its own score from it. Collect once, prove many: you keep one record, and each party reads it their own way.

The definition

Scoreable Proof describes how the record is organized, not a grade it hands out. Stated in its canonical form:

A governance history organized so any framework or algorithm — Law 25, AMF, OSFI E-23, NIST AI RMF, ISO 42001, EU AI Act, or a third party’s proprietary model — can compute its own score from it. Collect once, prove many.

The idea behind “collect once, prove many” is concrete. The same underlying facts — an AI inventory, a risk register, monitoring, a fix history — map to different vocabularies. The AMF calls it an SIA inventory with a cote de risque; OSFI E-23 calls it a model inventory with a model risk rating; the NIST-to-ISO 42001 crosswalk maps an AI-system-inventory control to specific ISO clauses. One clean history answers all of them, because each framework reads the facts through its own lens.

Who consumes it

Scoreable Proof exists for the parties who assess you — the relying parties. Each brings its own framework or model and reads your record for itself:

  • Auditors and regulators check whether you knew your AI, found the gaps, and addressed them — from records made at the time.
  • Insurers can price AI or cyber exposure from verifiable evidence, using their own models.
  • Lenders and boards can approve or extend credit on facts they check themselves, not claims they must take on trust.

Because the record is neutral and organized for scoring, none of them has to trust a number you supply. They compute their own.

Why Agentica never signs the score

This is the load-bearing limit. Agentica produces and holds the evidence; it never signs a relying party’s assessment. We are not a certification body, not a rating agency, and not the author of anyone else’s conclusion. Relying parties run their own reading on the record, with your authorization; the reference algorithms we offer are something a verifier may choose to run, never our signature.

That neutrality is exactly what makes the proof consumable by everyone. If Agentica graded you, every other assessor would have to trust our grade. Because we don’t, a bank, an insurer, and an auditor can each score the same record against their own standard. Agentica, an AI Governance company in Montréal, keeps that record tamper-proof and under independent custody — read-only, metadata and signals by default — so the facts hold up whoever reads them. Don’t trust the grade; verify the record.

Sources