Service 04 — AI Governance

Governance that runs, not governance that is documented.

An AI governance framework is the set of policies, controls and guardrails that make AI use lawful, auditable and trustworthy. We build the technical layer it depends on — and the evidence that shows the controls actually operate.

Compliance posture

Audit-ready evidence

NIST AI RMFAligned structure
ISO 42001Certifiable path
EnforcedOn the request path

Overview

Most governance programmes fail at the same point: they produce an excellent register of controls that do not run. We start from the enforcement layer and work outward — routing traffic through a governed path, then attaching inventory, classification, approval gates and evidence around controls that already operate.

The output is a control structure mapped to recognised frameworks — NIST AI RMF, ISO/IEC 42001, EU AI Act risk tiers — with a complete audit trail generated by the enforcement path itself rather than reconstructed afterwards.

Capabilities

AI system inventoryRisk classification & tieringControl framework adoptionPolicy encoding & enforcementEvaluation & approval gatesMonitoring & drift alertingIncident response runbooksAudit evidence collection

How we deliver

01

Inventory and risk tiering

You cannot govern what you have not catalogued. We build a living register of every AI system in use — owning team, purpose, model and version, data classification, environment, risk tier and monitoring configuration. Risk tiering then applies control intensity proportionally, which is what keeps the programme affordable and therefore actually followed.

02

Controls enforced on the request path

A policy that says 'do not send PII to external models' and nothing else is unenforceable. We encode policy as a decision on every request — allow, deny or redact — evaluated against data class, model and environment, with the decision logged alongside the request.

03

Evidence, not assurance

Auditors ask what happened, when, and under which policy. Every request produces a record containing its policy decision, redactions applied, model selected and outcome, retained to your compliance window. That record is the deliverable that distinguishes a governed system from a documented one.

04

Mapping to the frameworks you are assessed against

We structure around recognised frameworks rather than inventing a bespoke one — NIST AI RMF for risk management structure and vocabulary, ISO/IEC 42001 for certifiability, EU AI Act risk tiers for obligation severity — so your framework is legible to an auditor instead of proprietary.

What you get

  • A complete inventory of AI systems in use, including shadow AI.
  • Risk tiers with control intensity applied proportionally.
  • Policy enforced on every model request, not described in a document.
  • An audit trail generated by the enforcement path itself.
  • A control structure mapped to NIST AI RMF, ISO 42001 and EU AI Act tiers.

Frequently asked questions

What is an AI governance framework?

A structured system of policies, roles, controls and accountability mechanisms governing how an organisation develops, deploys and uses AI. The defining characteristic is repeatability: it specifies who decides, what evidence results, and what happens when a system leaves its approved bounds.

Do we need a governance platform and an AI gateway?

Usually both. The platform governs the programme — inventory, risk tiers, workflows, evidence. The gateway enforces controls on live traffic. One without the other produces either documentation without enforcement, or enforcement nobody maintains.

Which frameworks do you work with?

NIST AI RMF for risk structure, ISO/IEC 42001 for a certifiable management system, and EU AI Act risk tiers for obligation severity. We map your controls to these rather than building a bespoke framework.

How long does implementation take?

Inventory and risk classification are documentation work and can start immediately. The enforcement layer, including domain-specific detection models, is typically eight to sixteen weeks depending on how much labelled evaluation data we need to build from your traffic.

Does governance apply to third-party models we do not build?

Yes. Under the EU AI Act, obligations attach to deployers as well as providers, so consuming a third-party API does not transfer governance responsibility to the vendor.

Talk to an infrastructure architect

Bring your traffic profile, data constraints and cost targets. We will tell you what is feasible and what it would take.

Book a consultation