The motion

How a deal actually moves: the discovery questions that qualify a vendor, the CTA discipline that opens the next meeting, the two-sided shape of the whole motion — and the paragraph that reconciles Volume 1’s MGA ambition with Volume 2’s independence thesis.

Discovery: what to ask a vendor CTO

Four axes, half a page, everything you need to qualify the room. Each question is really a test of whether the denominator is tractable for this deployment.

Axis	Ask	What the answer tells you
Inference volume & attributability	“Roughly how many model calls a month? Can you attribute each one to a deployment, a customer, a model version?”	Whether a defined, attributable inference population exists — the wedge condition. High volume with clean attribution is the ideal profile. GROUNDED · Brochure
Audit pain	“What did your last audit or certification pursuit cost in engineering time? What did the assessor actually accept as evidence?”	Whether the compliance motion (the door-opener in 21-02) lands here — audit pain is the budget that already exists.
Insurance posture	“What does your E&O or cyber policy say about AI today? Seen any exclusions at renewal?”	Whether the insurance endgame is live for them. Exclusions are the demand signal, not a dead end. GROUNDED · Brochure
Deployment topology	“Where do the model calls flow — one gateway or scattered across services? Whose cloud, whose keys?”	Whether a governed passthrough is a day-one fit, and how big the governed scope — the only scope you make claims about — will actually be. GROUNDED · Labs

A CTO who answers all four crisply is a design-partner profile. A CTO who can’t answer the first one is the thesis walking: they are carrying unpriced, uncounted exposure — which is a finding, and findings open meetings.

CTA discipline

One primary, one secondary, nothing else. GROUNDED · Labs

• Primary — “See it in action.” Low commitment, first-person, concrete.
• Secondary — “Book a 30-minute assessment.” The scoped, named next step; also the pricing deflection (03 The pricing conversation).
• Deprecated — do not use: “Book a demo,” “Schedule a conversation,” and personal scheduling links. Vague CTAs read as vague products, and the Floor’s judge scores them accordingly.

The two-sided motion

The same attestation record is the product on both sides of the trade:

• Sell to AI vendors and deployers: enforcement, evidence, certification acceleration — and eventually their attested history as premium leverage.
• Support the market side: syndicates, reinsurers, and brokers who consume that same evidence as parametric-trigger input and accumulation signal.

You are never choosing between the sides; every attested deployment strengthens the market-side story, and every market-side conversation raises the value of being an attested deployment. That is Volume 1’s flywheel with the evidence layer at the hub. GROUNDED · Brochure

“Volume 1 said Glacis becomes an MGA. Which is it?”

The reconciliation, honestly: Volume 1’s three streams included a producer/MGA path in which Glacis itself underwrites. The strategy has evolved. The current position is the brochure’s: Glacis does not hold risk — “the moment Glacis holds a book, every carrier treats attestation receipts as a competitor’s proprietary format — and the standard collapses.” GROUNDED · Brochure The economics survive the change: Glacis earns per attested inference population as the evidence provider, while the syndicate holds the risk and keeps the underwriting profit. The flywheel still turns and the loss ratio is still its hub — what changed is Glacis’s seat at the table: the evidence layer, not the paper. Volume 1’s licensing knowledge remains what it always was — market literacy, and the reason you can hold this conversation at all. See 19-01 The independence thesis.

Next

→ 03 The pricing conversation

Drill this page →17 bank questions stand behind what you just read. Check it while it’s warm.