AI Use Assessment: What Your AI Can Evidence · Triodian

Assessment · available now

Find out what your AI, and your providers' AI, can actually evidence.

The AI Use Assessment registers every place AI acts for you: in your own teams, inside your providers and in the platforms behind them. It asks the same seven questions of each, and records what evidence can be produced and independently checked.

Available nowFixed feeThree to four weeksRun through your counselNo customer data

The question it answers

At each boundary in your chain, what evidence of AI behaviour can be produced, measured against your board's appetite?

How it runs

Four steps, the same for every arrangement.

01

Confirm the set

A representative set of arrangements, agreed with the CRO or GC. Not the whole estate.

02

Map the AI

Where AI acts, what authority it holds, and who stands behind it. We start from your own inventory.

03

Classify and test

The consequence of each use, and the evidence obtainable at each boundary.

04

Questions to counsel

Legal questions go to counsel in writing. Counsel sees every draft first.

The evidence ladder

The evidence asked for rises with what the AI can do.

Each authority level has a default rung. The consequence of failure and the rights you hold can move it up or down.

Authority delegated to the AIDefault evidence rungWhat you hold
ProductivityDisclosureA statement of where and how AI is used
AnalyticalAttestationA signed statement that stated controls operate
DecisionEvent notificationNotice when defined thresholds are crossed
TransactionalAuditA right to have the evidence examined
AgenticContinuous evidenceA periodic, aggregate measure against appetite
ActuatingDeterministic controlA control at the point of action, with a signed record

What you provide, and what you get

Your inventory and your agreements. No data.

Through counsel, we need your current AI inventory and the agreement for each arrangement in scope. No customer data, no individual decisions, no system access and no provider involvement.

Triodian produces

  • The AI use register, including AI inside providers and the platforms behind them
  • The evidence obtainable at each boundary
  • A finding per arrangement: the gap and its owner, or no reasonable reach
  • A technical schedule beside counsel's flow-down clauses
  • Tolerance and trigger design, and the shape of a control library for your own AI
  • A technical annex, in Triodian's name

Counsel produces

  • A reach determination for each arrangement type
  • Grading: requirement, expectation or better view
  • The accountability note and the privilege position
  • A flow-down clause schedule
  • The legal half of the board paper

Who does what

No party marks its own work.

PartyRoleWhere it stops
YouThe CRO commissions the work. The GC instructs counsel and sets the brief. The board owns appetite.Your AI, governance and safety teams own the gaps handed back to them.
Board risk adviserFacilitates the appetite sessions and writes the board paper. Optional: we can work with yours.Does not perform the assessment.
CounselEngaged before any finding is written. Owns every legal call and the privilege position.Sees every draft first.
TriodianOwns the method and the technical result. Builds the register, runs the assessment, operates the apps.Gives no legal advice. Presents no product.
Independent checkerChecks the assessment for counsel: is it sound, and are its findings supported?Engaged separately from Triodian. Builds nothing.

We work with large law firms and board risk advisers in alliance with Triodian, or with yours.

Where it stops

The work can stop at the findings.

No legal advice

Every legal determination is yours and your counsel's. Legal questions go to counsel in writing.

No product presented

The measure is your supervisor's expectations and your board's appetite, not any product.

Aggregate evidence only

No customer data, no individual decisions, no system access, no access to a provider's model.

"No reasonable reach" is a finding

Where you hold no rights, that is recorded honestly. The work can stop at the findings.

After the finding

Who closes a gap, and with what, is your separate decision.

Whichever route you choose, the test is the same: can the evidence be produced, and can someone independent check it?

Terms in the agreement

Evidence terms at the next contracting point. Counsel drafts them, with the technical schedule beside the clauses.

The provider's own solution

A provider may already have a control or record that closes its gap. The finding says what it must show.

Tools you already have

An existing governance, security or board tool may cover a gap inside your own estate.

Triodian, if you ask

A boundary control and evidence design, as a separate engagement decided on the findings.

Evidence rungWhat it has to showTriodian route, if chosenHorizon
DisclosureWhere AI acts and what authority it holdsThe AI use register, from this assessmentAvailable now
AttestationThat an output met a declared policyCertifiable Conformance PackIn build · 2027
Event notificationThat a set of decisions moved outside appetiteAggregate Drift ServiceIn build · mid-2027
AuditWhat was checked, against which policy versionCompliance token and ledgerIn build · 2027
Continuous evidenceThat the check still holds after the model changesHarness-as-AssuranceIn build
Deterministic controlThat a forbidden action could not be takenBounded Command, then the hardware applianceIn build · then 2028

The app

Run in the AI Use Assessment app.

Eight phases, each with a gate; anything produced before its gate is met is marked draft. One register entry per place AI acts, the seven questions, one obtainability test per evidence rung, legal questions routed to counsel, and an independent check before the findings are handed over.

Available now. The worked example uses a fictional insurer and its providers.

Get in touch

Thirty minutes to test whether the question is live for you.

Whether AI sits inside a material provider, and whether a review, a renewal or a deadline is coming.

Book a 30-minute conversation →