AI governance · Built to be run by your people
AI Governance Office
A working AI governance function stood up inside your organisation: a policy that routes, intake and risk tiering mapped to the EU AI Act, approvals proportionate to risk, controls with evidence behind them and board reporting that keeps pace with adoption. Fixed fee programmes from £30,000 + VAT, piloted on your real use cases before handover.
The Office at a glance
Definition
What is an AI governance office?
An AI governance office is the standing function that governs how an organisation adopts AI: the policy that routes decisions, the intake that catches new use cases, risk tiering mapped to the EU AI Act, approvals proportionate to risk, controls with evidence behind them, and reporting a board can hold. It is an operating capability rather than a document, which is why a policy PDF and a quarterly ethics committee keep failing at the job.
Intelance designs and stands one up as a fixed fee programme from £30,000 + VAT, mapped to the EU AI Act and ISO/IEC 42001 and built to be run by your people. Retained office models afterwards run from £8,000 + VAT per month and are entirely optional.
The build starts from evidence: your real inventory, your real tiers, your real shadow AI. The AI Governance QuickScan from £15,000 + VAT provides exactly that and prices the Office precisely, and the tiering runs against the Act as amended in June 2026. Our guide to building an AI governance office covers the anatomy, and the wider practice is described under AI governance consulting.
A policy that routes
Decisions flow to the right place by design, rather than to whoever shouts.
Intake that catches everything
New use cases enter one door, including the ones that used to sneak in.
Approvals sized to risk
A chatbot pilot and a credit model do not queue in the same line.
A board that knows
Adoption, risk and incidents reported in business terms, on a rhythm.
The anatomy
Six parts of a working AI governance office
Committees and policies fail because they carry one or two of these. A working office carries all six, and each card shows the output it leaves running.
The policy that routes
A short policy that tells anyone with an AI idea exactly where to take it and what happens next, rather than a values statement nobody opens.
The outputOne page people follow, because following it is faster.Intake and tiering
Every new use case enters through one door and is tiered against the EU AI Act and your own risk appetite within days.
The outputA register that stays complete, because intake feeds it automatically.Proportionate approvals
Low risk moves fast, high risk gets scrutiny, and the difference is written down. Blanket caution is how shadow AI wins.
The outputApproval routes sized to risk, with turnaround times people can plan around.Controls and evidence
The controls that sit between users and models: access, data boundaries, human oversight and logging.
The outputControls that generate their own evidence as they run.The forum and rhythm
A standing forum at delivery pace that decides, records and unblocks, instead of a quarterly committee that watches the backlog grow.
The outputA decision log the whole organisation can rely on.Board reporting
What is running, what is queued, what went wrong and what the EU AI Act clock says, in business terms.
The outputA board pack that survives scrutiny, quarter after quarter.How it works
How the office is built and run
Governance functions fail as documents and succeed as habits. The Office is proven on live use cases before it is rolled out.
From the evidence
The Office is designed from your real inventory, tiers and shadow AI, usually straight from the QuickScan. Fresh discovery runs inside the programme where needed.
The working parts
Policy, intake, tiering, approval routes, controls and the register, built with the people who will run them.
Real use cases
The Office runs on live decisions from your actual AI pipeline before rollout. A route that fails a real use case is redesigned here.
The rhythm
The forum runs at delivery pace, the register stays current, and the first board pack lands.
Run it yourself or retain us to run it
Your people run it from handover, or the retained office runs it with you from £8,000 + VAT per month. Either way, the exit is designed in.
Most builds run over a few months from evidence to handover, with the working parts standing inside the first six weeks. Larger and multi-entity estates take longer. The QuickScan first shortens everything.
Published fees
The services around the Office
Every route in and out of the build is priced here, before you ask.
AI Governance QuickScan
The evidence base the Office is built from: every use case found including shadow AI, tiered against the EU AI Act, in two weeks. It prices the build precisely.Best whenNobody yet knows what AI the organisation runs. FeeFrom £15,000 + VAT2 weeks 02
AI Risk-as-a-Service
The governance run for you month by month: intake worked, tiers kept current, the board kept informed, without building the standing function first.Best whenYou need governing now and building later. FeeFrom £8,000 + VAT / month 03ISO/IEC 42001 Implementation
The AI management system standard implemented on top of the Office, for customers and regulators who want certified proof.Best whenThe ask has moved from assurances to certification. FeeAIMS implementationPriced by scope 04
EU AI Act Readiness
Exposure mapped against the Act in depth: roles, obligations, documentation and the route to compliance before enforcement asks first.Best whenThe Act specifically is the board’s worry. FeeReadiness assessmentPriced by scope
The alternatives
This route against the alternatives
Three ways organisations respond to an AI governance mandate. The difference shows the first time a real use case arrives.
A responsible AI policy on the intranet, unread since the day it was published
Nothing happens. Nobody knows the policy applies to them
Nothing, until something goes wrong
Organisations that need to say something exists
A quarterly committee of senior people with no intake, no register and no controls
It waits a quarter for an opinion, so teams stop asking
Minutes
Genuinely slow, small AI adoption
A standing function with policy, intake, tiering, approvals, controls and reporting, from £30,000 + VAT
Through the door, tiered in days, approved on a route sized to its risk
Adoption, risk and the compliance clock, in business terms, on a rhythm
Organisations where AI is moving faster than anyone is governing it
Scope
What is covered and what is not
You get
A routed policy, an intake door and a register that stays complete because intake feeds it.
Risk tiering mapped to the EU AI Act and ISO/IEC 42001, with the post-Omnibus duties assigned to owners.
Approval routes proportionate to risk, with turnaround times teams can plan around.
Controls with evidence behind them, piloted on your real use cases before rollout.
Board reporting on a rhythm, plus a development path for the function itself.
This is not
A policy exercise. The Office is an operating function. The documents are its record.
A brake on adoption. Proportionate approval exists so the answer is usually yes, safely and quickly.
A committee. The forum decides and unblocks at delivery pace. It does not meet quarterly to watch the backlog grow.
A dependency. Built to be run by your people, with the retained model entirely optional and the exit designed in.
Where to start
Typical starting positions
Four situations we are usually called into. If none of them fit, speak to us and we will point you at the right first step.
Why Intelance
The standards we hold ourselves to
Governance that lets you approve AI quickly and safely
The Office exists so the organisation can adopt AI faster, with control. Proportionate routes are the whole design, because blanket caution is how shadow AI wins.
Built from evidence
The design starts from your real inventory, tiers and shadow AI, usually straight from the QuickScan, never from a template.
Piloted on real use cases
No route survives to rollout until it has carried live decisions from your actual AI pipeline. Routes that fail the pilot are redesigned there.
A published fixed fee
Programmes from £30,000 + VAT by scope, agreed in writing before work starts, with the drivers of the fee stated plainly.
Yours by design
Your people run the Office from handover. The retained model from £8,000 + VAT per month exists only if you want it, and you can move either way later.
Who delivers this
Every Office is built by a senior AI governance lead with the people who will run it. Nothing is offshored and nothing is templated.
AI governance guides
All insights
AI GovernanceHow to build an AI governance office that works
The build guide for the function that makes AI governance real: the mandate, the staffing maths, the operating rhythm, the first sixty days, and the five signals that show it is working rather than performing.
9 July 2026 · 6 min read Read the guide
AI GovernanceWhat is AI governance? The seven questions boards should ask
9 July 2026 · 9 min read
AI GovernanceEnterprise AI governance in 2026: a practical blueprint for CIOs and boards
9 July 2026 · 8 min read
AI GovernanceAI governance vs AI assurance: what is the difference?
9 July 2026 · 6 min readReady to talk?
Tell us what is driving the mandate
Thirty minutes with the person who would build your Office: what the board asked for, when, and what evidence already exists. If the QuickScan has run, the build is already priced. If not, we will tell you which order saves money.
FAQs
Questions about the AI Governance Office
What is an AI governance office?
The standing function that governs how an organisation adopts AI: a policy that routes decisions, intake that catches new use cases, risk tiering mapped to the EU AI Act, approvals proportionate to risk, controls with evidence behind them, and board reporting on a rhythm. It is an operating capability rather than a document.
How much does it cost to implement AI governance?
With Intelance, Office build programmes run from £30,000 + VAT, fixed by scope and agreed before work starts. The fee is driven by how many AI systems you run, the sectors you operate in and what the QuickScan found. Retained office models afterwards run from £8,000 + VAT per month and are entirely optional.
How long does the build take?
Most builds run over a few months from evidence to handover, with the working parts standing inside the first six weeks, a pilot on live use cases, and the rhythm embedding after that. Larger and multi-entity estates take longer, and we say so at the start.
Is this the same as an AI governance framework?
The Office is the framework made operational. A framework on paper describes tiers, roles and controls. The Office is those same parts running: intake worked daily, approvals turned around on published times and a register that stays complete. The framework documents become the record of a working function.
What is an AI governance operating model?
The design underneath the Office: who holds which AI decision rights, which forum decides at what threshold, how intake, tiering and approval interlock, and what the board sees. The build makes those choices explicitly with you, then pilots them on real use cases before anything is rolled out.
Do we need the QuickScan first?
No, but it usually saves money. The Office must be designed from a real inventory either way. The QuickScan from £15,000 + VAT delivers that in two weeks and prices the build precisely. Without it, discovery runs inside the programme instead.
Who runs the Office after handover?
Your people, by design: they run the pilot with us, chair the forum before we leave and own the register from day one. Where capacity is the constraint, the retained office runs it with you from £8,000 + VAT per month, and you can move either way later.
How does the Office map to the EU AI Act?
Risk tiering follows the Act’s categories, the register records the roles and obligations each use case attracts, and the duties as amended, including the 2 December 2027 Annex III obligations, are assigned to named owners with dates. The board pack carries the compliance clock, so nothing arrives as a surprise.
What does the retained office include?
The rhythm run with you month by month: intake worked, new use cases tiered, approvals administered, the register and controls kept current, and the board pack produced. It is the same service described under AI Risk-as-a-Service, from £8,000 + VAT per month, with the exit designed in.