Personas and stakeholders
Seven rooms you will be in, what each person in them is actually protecting, and the one grounded answer to their sharpest objection. The motion is two-sided (see 02 The motion): the first three personas buy the evidence layer, the next three consume it, and the seventh signs the check.
The buyer side (AI vendors and deployers)
Section titled “The buyer side (AI vendors and deployers)”| Persona | Cares about | Sharpest objection | The grounded answer |
|---|---|---|---|
| Healthcare CIO / CTO (deployer) | Patient safety, PHI exposure, board reporting | “You’re going to see our patient data.” | Zero egress: raw prompts and responses never leave your environment — only hashes and signed metadata cross the trust boundary. GROUNDED · Labs |
| CISO / governance buyer | Board evidence, audit prep, tamper-evidence | “Who verifies the verifier?” | The Notary is independent by construction and OVERT is open — any auditor can verify a record offline, without trusting Glacis or the vendor. GROUNDED · Labs |
| Compliance / regulatory officer | ISO 42001, FDA/MHRA/EU AI Act artifacts | “We’re already doing this with a GRC tool.” | GRC tools collect documents; this generates runtime evidence continuously — and point-in-time assessments “expire the moment a model updates.” GROUNDED · Brochure |
Two more objections worth pre-loading for the CIO/CTO room: latency (the policy hot path is deterministic and engineered for production speed — no LLM calls inside the runtime GROUNDED · Labs) and “we already log everything” (logs are mutable and self-maintained; attestations are witnessed — “Configuration is a promise. Execution is a fact.” GROUNDED · Labs).
The market side (who consumes the evidence)
Section titled “The market side (who consumes the evidence)”| Persona | Cares about | Sharpest objection | The grounded answer |
|---|---|---|---|
| Lloyd’s underwriter / actuary | Pricability, basis risk, data credibility | “Your denominator is vendor-reported — a management assertion, not evidence.” | Concede the sharpness, then walk the denominator trust ladder: OVERT classifies denominator source explicitly, and only Notary-bounded counts are pitched as underwriting-grade. GROUNDED · Labs |
| Reinsurer / accumulation manager | Correlated loss from foundation-model updates | “One data vendor can’t see cross-portfolio correlation.” | Portfolio-level behavioral-correlation visibility is exactly the proposed accumulation demonstration — version-change detection catches the same upstream update across deployments. GROUNDED · Brochure (Proposed cohort deliverable — hedge it.) |
| Broker / MGA partner | A sellable product, commission, capacity | “Carriers are filing blanket AI exclusions — there’s no paper to place.” | Exclusions are the demand signal: insureds facing them need the evidence that makes them insurable, and “credit for cleaner attestation records” is a pricing story a broker can defend. GROUNDED · Brochure |
The underwriter’s objection (a) — “no loss history, can’t price” — gets the structural answer from 20-00: parametric triggers price from attestation data alone, and nobody has loss history because nobody had a denominator. In an underwriter’s own recorded words: “They never come to me with a denominator… You cannot price.” GROUNDED · KB
The check-signer
Section titled “The check-signer”| Persona | Cares about | Sharpest objection | The grounded answer |
|---|---|---|---|
| Procurement / CFO (at the AI vendor) | Cost vs revenue, lock-in, ROI | “This is a cost center — compliance can wait.” | It’s revenue enablement: enterprise buyers increasingly require proof of AI governance as a purchase condition, and certification is itself a sales gate the vendor is already paying for. GROUNDED · KB |
Their lock-in objection has a clean architectural answer: OVERT is open, so receipts are verifiable by any party without Glacis — neutrality is structural, not promised. Their ROI demand gets the honest treatment in 03 The pricing conversation — note that premium-relief outcomes for customers are not yet demonstrated (VERIFY; never imply otherwise).
Reading the table like a producer
Section titled “Reading the table like a producer”Every grounded answer above is a tier-i or hedged tier-ii claim — that is what makes it usable in a live room. If an answer you’re about to give doesn’t appear on this page or reduce to the claim taxonomy, hedge it or drop it.
Drill this page →17 bank questions stand behind what you just read. Check it while it’s warm.