Assurance

Evidence that your controls hold.

Claiming you have AI governance and being able to prove it are different things. Assurance is independent verification: that vendor claims are accurate, that your models behave as described, and that the controls you've put in place are working in practice, not just on paper.

The Distinction

Policy tells you what should happen. Assurance shows what does.

Policy tells you what should happen. Assurance shows what does.

Governance documents describe intended behaviour. Assurance examines actual behaviour. It tests whether the vendor's claims about their model match how the model performs on your data. It checks whether your operational controls are followed in practice. It produces the documentation a regulator, general counsel, or board audit committee can evaluate.

Strategy consulting firms hand you a deck. Meninge installs a living scenario portfolio your leadership team continues to use. The scenarios get refined as conditions change. The work stays

useful.

Most mid-market organisations have some governance documentation. Far fewer have independent evidence that it holds. That gap is where liability actually lives.

Strategy consulting firms hand you a deck. Meninge installs a living scenario portfolio your leadership team continues to use. The scenarios get refined as conditions change. The work stays

useful.

Governance without assurance is intention. Assurance is the evidence.

Scope of Assurance Work

Three areas. All of them verifiable.

Three areas. All of them verifiable.

Vendor & Model Claims

AI vendors make claims about how their models work, what data they were trained on, and how they perform across different populations. Independent review tests those claims against your actual operating environment. We examine model cards, audit documentation, contractual representations, and performance outputs to determine where vendor claims hold and where they don't.

What you walk away with

Vendor documentation review: model cards, data sheets, audit reports

Gap analysis between vendor claims and observed model behaviour

Contractual representation assessment: what the vendor has committed to in writing

Risk rating for each gap with recommended remediation

Summary suitable for General Counsel and procurement review

Model Behaviour

How a model behaves in a controlled test environment and how it behaves on your data can differ significantly. We test the model's outputs against your use cases, examine performance across relevant demographic groups where data is available, and identify drift, inconsistency, or behaviour that does not match the stated design.ors make claims about how their models work, what data they were trained on, and how they perform across different populations. Independent review tests those claims against your actual operating environment. We examine model cards, audit documentation, contractual representations, and performance outputs to determine where vendor claims hold and where they don't.

What you walk away with

Output consistency assessment across operational conditions

Fairness and bias review where decision-impacting outputs are involved

Performance gap analysis between vendor benchmarks and your operational data

Drift assessment for models that have been in production for 12 months or more

Written findings with evidence for each finding

Operational Guardrails

Controls that exist on paper but are not followed in practice provide no actual protection. We review whether your governance processes are being applied, whether staff are following documented protocols, and whether your escalation paths function as designed. The output is documentation a regulator would accept as evidence of operational governance.

What you walk away with

Process walk-through with operational staff to verify documentation reflects practice

Escalation path verification: does the documented path match the actual path

Oversight mechanism review: who reviews what, how often, and with what authority

Gap report with prioritised remediation steps

Board-ready assurance summary

Typical Buyers

Who commissions assurance work.

Who commissions assurance work.

General Counsel / CCO

When documentation exists but operational evidence does not

Martin is on every call, in every session, writing every

deliverable. No junior handoff. The person with the

credentials does the work.

Procurement Teams

Before vendor claims become your liability.

Buying an AI tool based on vendor claims is a liability if those claims are not independently verified. Procurement teams use assurance work to close the gap between what a vendor says their model does and what it actually does in your environment.

Boards & Audit Committees

Something substantive to evaluate — not just a policy to accept.

Boards are increasingly accountable for AI governance. An independent assurance review gives the board something substantive to evaluate, not just a policy document to accept.

CEOs Post-Incident

After something has gone wrong — and before it goes further.

After an AI-related incident or near-miss, assurance work establishes what the controls were, whether they were followed, and what failed. It is also the basis for remediation.

Most firms give you a report. Meninge gives you a position you can defend.

30 minutes, on the record, no obligation. We'll tell you whether we're the right fit before you spend a dollar.

meninge

AI strategy and governance for mid-market organisations.
Ottawa — serving Canada and the United States.

Fineprint

© 2026 Meninge

meninge

AI strategy and governance for mid-market organisations.
Ottawa — serving Canada and the United States.

Fineprint

© 2026 Meninge

meninge

AI strategy and governance for mid-market organisations.
Ottawa — serving Canada and the United States.

Fineprint

© 2026 Meninge