SystemCompliance & Regulatory

AI Governance

AI Governance & Model Inventory

Every AI system the insurer uses inventoried, tested and documented, ready for the regulator’s first question.

Film coming soonSee it working, liveWe’ll walk you through AI Governance on your own documents.
SystemA crew of agents runs a whole process; people decide.
6specialist agents
5kinds of input
Soonfilm in production
The problem

Why it matters

Models sit in pricing, claims, fraud and vendor tools with no single inventory; when an examiner asks how each was tested, teams spend weeks chasing owners and contracts.

What it does

How the solution handles it

Agents find every AI system in use, write a cited model card, run outcome tests across groups, collect vendor evidence, and map each system to the NAIC bulletin, Colorado and other rules. A compliance officer reviews the gaps and signs the attestation.

How it works

How an AI system moves

Six specialist agents find, document, test and map every AI system to the rules; a compliance officer attests and signs.

What comes in
Systems in4 sources · models, vendor contracts, owners, policies
Agents at work
Model findersystems and vendors
Then
Model card writerpurpose, data, owner
Then
Bias test runneroutcomes by group
Vendor evidence checkerthird-party models
Then
Rule mapperbulletin to control
Then
Evidence pack builderfor exams, cited
A person decides
Compliance officerattests and signs
What comes out
AI system inventory
Exam-ready evidence
Signed attestation
In and out

What it reads, and what it hands back.

What goes in

  • Model and system registers
  • Vendor contracts and model documentation
  • Test and monitoring results
  • AI policies and board minutes
  • Bulletins and rules (NAIC, Colorado, NY, EU)

What comes out

  • AI system inventory with owners
  • Model card per system, cited
  • Exam-ready evidence pack
  • Signed attestation

Who uses it

COCompliance officersMRModel risk teamsDSData science leadsIAInternal audit
What it changes

The difference, in numbers.

Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.

target
2days
to answer a regulator’s AI questionnaire with cited evidence
By hand3–6 weeks
With agents≈ 2 days
target
100%
of AI systems with an owner, purpose and last test date
every system accounted for
estimated
4×
the models the same team reviews each quarter
TodayWith agents

Sources: NAIC Model Bulletin on AI Systems (Dec 2023; 25 states + DC by Aug 2026) · Colorado 3 CCR 702-10 · “Target” = design goal, measured in the live solution · “Typical” = typical industry figure · “Estimated” = our estimate

Built on the engine

6 specialist agents. One person decides.

Model finderin-house, vendor and embedded AI
Model card writerone cited card per system
Bias test runneroutcome tests across groups
Vendor evidence checkercontracts, docs and test reports
Rule mapperNAIC, Colorado, NY and EU AI Act
Evidence pack builderinventory and evidence on request
Compliance officerattests and signs

Build this
for your team.

We’ll show AI Governance running on your own documents.