Episode

AI governance soothes the board. It doesn’t steer the lift.

Rory O’Keeffe, founder of RMOK Legal and former Director of Legal Services at Accenture, on the distance between AI policy and practice, why signature is the starting gun, and the one question leaders should ask their teams this week.

Published

16 November 2026

Duration

38 minutes

Guest

About the speaker

Rory O’Keeffe is the founder of RMOK Legal and a former Director of Legal Services at Accenture and partner at Matheson. Over more than 20 years, he has negotiated the technology deals where risk actually gets decided. He is co-author of AI Advantage, and looks at AI governance from where the contract meets the people using the tools.

Episode guest

In this episode

Rory describes the gap as distance: the people writing the rules sit three floors up and six months away from the people using the tools. On paper, there’s a policy and a steering committee. In reality, there’s a sales director at eleven at night, pasting the client’s pricing into a chatbot to get a proposal out by nine. It works, so it happens again.


The bigger blind spot is AI nobody chose: a feature quietly switched on in software you bought years ago, like the note-taker summarising client calls or the CRM that now drafts emails. There’s no new contract and no new review. Your framework approved the vendor as it was, not what it has become.


“A surprising amount of AI governance works like the close-door button in a lift. It makes you feel in control. It soothes the board, but it doesn’t steer the lift.” - Rory O’Keeffe


His test is three questions. Can anyone name the person who owns AI risk? When was your approved-tools list last updated? When did someone last check what your existing software can now do? If the answers don’t come easily, don’t rewrite the policy. Look at the workflow, because that’s where governance lives.


Signature is the starting gun, not the finish line.


Is risk decided in the contract or after it’s signed? Both. The contract decides who pays when something goes wrong. What happens after signature decides whether it goes wrong.

Teams fight for weeks over a liability cap, then file the contract and never open it again. After that, the real decisions are small and almost invisible: a scope change agreed on a Friday afternoon, updated terms accepted with a click, a renewal that rolls over because nobody raised questions in time. You have the most leverage the day before you sign, and a second chance before renewal. Most companies waste both.


AI makes this harder. Most AI contracts are standard software contracts with “AI” added to the definitions, written for software that behaves the same way every time. The model you buy on day one isn’t the one you’ll have on day 90.


“Contracts don’t manage themselves. They just sit there being right, waiting for someone to notice.” - Rory O’Keeffe


Stop asking if they use AI. Ask them to show you.


What should teams start on now? Evidence. Sooner or later, someone will say “show me”: what you approved, who checked it, and what you did when it went wrong. “Dave was at the meeting, and Dave’s left” isn’t evidence. It’s folklore.


For leaders, Rory suggests a small change of wording. Don’t ask your team whether they use AI, because everyone gives the right answer. Ask them to show you how they used it last week. You’ll find the workaround that saves six hours a week, and the customer data pasted where it shouldn’t be. The point isn’t to catch anyone. It’s to find the workarounds good enough to become the official way.


“Ask your team to show you how they used AI last week. In an hour, you’ll learn more about your real AI risk than from any policy review.” - Rory O’Keeffe

More

More from the series

Worth your time – security, privacy, and resilience leaders across Europe.

When Controls Meet RealityOperational Resilience, DORA, Risk Management, Financial Service
FriSept 2026

When Controls Meet Reality

A framework that reads well for auditors can still fail when pressure hits. In this episode of Montro In Practice, Ankur Arora sits down with Emmanuel Ranchin, VP of Information Protection Governance at State Street, on the gap between documented controls and lived ones: why compliant isn't the same as resilient, how big institutions break where silos meet, and what DORA really asks of the board.

Watch now
View all

Don't miss the next conversation

Get new Expert Series episodes and governance insights in your inbox.

Browse episodes