Relay & Co. · Primer

What is AI governance?

AI governance is how an organisation makes its AI systems safe, compliant and accountable — the policies, processes and controls that turn ethical principles into practice.

Read the practical guides The building blocks
01 — Anatomy

The building blocks

Five components. Everything else is detail, sequencing, or someone's product.

Oversight

A named person or team accountable for AI risk. If the answer to “who owns this?” is a committee, the answer is nobody.

Inventory

Knowing which AI systems you actually use — including the shadow AI nobody put through procurement.

Risk assessment

Identifying what could go wrong, how likely it is, how bad it would be — and how you would know it had happened.

Policies

Clear rules on what is allowed, what needs approval and what is forbidden. One page beats forty nobody reads.

Controls

The practical guardrails that enforce the rules: logging, human handoff, access limits, retention, review gates.

Where to start

The jargon-free starter guide walks through the first four moves in order.

Guide
02 — Timing

Why it is urgent

Three forces are converging, and they are not waiting for your roadmap.

Regulation

The EU AI Act and a widening set of sector rules have moved from proposal to obligation. Deadlines land whether or not the inventory exists.

Board-level liability

Accountability for AI harm is climbing the org chart. Directors are being asked what they knew, when, and what they did about it.

Speed of adoption

AI tools spread inside organisations faster than any governance process moves. By the time policy is drafted, the tools are already in use.

03 — Standards

Frameworks that help

You do not have to invent a system. Three references cover most of the ground — pick one as your spine and map the rest to it.

Regulation

EU AI Act

The risk-based regulatory framework: prohibited uses at one end, minimal-risk applications at the other, with real obligations in between.

Plain-English guide
Standard

ISO 42001

The AI management-system standard — a certifiable structure for governing AI that auditors increasingly recognise as the practical route.

Where it fits
Framework

NIST AI RMF

The U.S. risk-management framework: voluntary, flexible, and a useful common language for teams that need to describe AI risk consistently.

Terms explained

Governance is less about technology than leadership hygiene

Treat it as a business discipline — ownership, records, review — rather than a technical programme, and it becomes something an organisation can actually sustain. Read the practical guides →