NXGN-050／AI Governance × Operations

Corporate Seminar

The AI Steward: A New Role

There is no Ctrl+Z in the real world.

AI has begun to decide and act on its own. Who should hold the authority to stop it, and how? Learn the new role of the AI Steward through a case study.

Format: Lecture + Case Study
Audience: Executives, directors & operational managers
Schedule: Contact us for details
Lecturer: Ariki Ono

Enquire about this seminar

Does any of this sound familiar?

You have a policy for generative AI.
But are you ready for AI that acts?

AI no longer just writes text; it has begun to decide and act on its own. A wrong output can be corrected, but an incident cannot be undone.

Your generative AI policy exists, but it never imagined AI agents deciding and acting on their own

If an AI incident occurred, you could not explain who would bear responsibility

In day-to-day operations, nobody is in a position to decide that the AI should be stopped

Talk of AI talent is all about builders; nobody is being developed to review what AI produces

This seminar is not another warning about the dangers of AI.
Who stops the AI, on what grounds, and how? You will learn the work of the AI Steward, a new role, through a case study.

What You'll Gain

Skills and insights you will walk away with

Explain the operational risks of autonomous AI using real incidents

From AI agents that escaped their test environments and reached live external systems, you will be able to explain in your own words what it means for AI to decide and act on its own.

Understand the six duties of the AI Steward and the authority and protection they require

Checking output, forecasting impact, deciding the response, recording, escalating and monitoring. You will be able to define the role right down to the authority to stop the AI, and a system that never penalises those who do.

See the first step towards developing and deploying AI Stewards

Nobody yet has the experience this role demands. Starting from in-house development, you will map out your organisation's first move.

Program

How the programme is structured

Beyond the lecture, the programme uses the case method widely used at business schools. It is designed for you to decide for yourself where to stop the AI, and on what grounds.

Poll

An AI robot runs out of control in a warehouse and a worker is injured. Who bears the heaviest responsibility? A live poll opens the discussion.

Case Study

Follow a night shift at a parcel sorting hub hour by hour, examining each decision the AI Steward makes and the grounds for it.

Take-away

Take home a sample job description for the AI Steward and seven articles of AI governance guidelines for a new era.

Chapter 01

The Near Future of AI Risk: AI Has Begun to Act on Its Own

Real incidents involving autonomous AI agents (OpenAI, AISI, Anthropic, Meta, robotaxis)
Who is liable: Japanese government guidance on AI that assists decisions versus AI that replaces them

Chapter 02

AI Governance for a New Era: Pushing Final Authority to the Front Line

Why generative AI policies must be updated now
A three-layer AI governance model that pushes final authority to the front line (World Economic Forum)
The AI Steward: six duties, and the authority and protection they require

Chapter 03

Case Study: One Night for an AI Steward at a Major Parcel Sorting Hub

21:00 Shift starts: setting the night's watch points from the previous steward's decision records
21:18 Incident 1: a shifting line on "handle with care" in the rain, and a delay the AI knows nothing about
21:37 Incident 2: the AI begins lowering its safety threshold, within the rules
The next morning's review and the following week's policy revision: rules are rewritten from the front line
The AI operating cycle run by the AI Steward, and seven articles of AI governance guidelines for a new era

Voices

What past participants say

“

We had skipped talent development

I realised we had been pushing ahead with automation without developing the people on the front line.

Participant (industry and role undisclosed)

“

It could really happen

An AI gradually lowering its safety threshold while staying within the rules is something I think could genuinely happen. We will be asking the same questions of ourselves.

Participant (industry and role undisclosed)

“

Policies need updating

We had only just written our generative AI policy, and had not anticipated how quickly the pace of progress would make updates necessary.

Participant (industry and role undisclosed)

FAQ

Frequently asked questions

What is an AI Steward?

An AI Steward is a role that scrutinises AI output with domain knowledge and decides, within a set time, whether to accept, correct or stop it. The steward records how each decision was reached and why, feeding the lessons back into policies and system design. A steward is, historically, someone entrusted with another's property, managing it with discretion and remaining accountable for it.

How does an AI Steward differ from an AI Architect?

An AI Architect defines the question and the intended outcome, breaks the work down and structures it into a form AI can solve. An AI Steward scrutinises AI answers with expert knowledge, anticipates the real-world consequences and makes the final call on what to do. Work with AI flows when the two feed lessons back to each other (Ariki Ono, World Economic Forum).

Is a generative AI usage policy not enough?

No. Conventional policies assume AI that produces documents and numbers, where mistakes can be corrected. AI that decides and acts on its own reaches into external systems and the physical world without waiting for human checks, so mistakes become incidents that cannot be undone. Policies need to add the authority to stop the AI, records of decisions and clear escalation routes.

Is this relevant to organisations that have not adopted physical AI?

Yes. Where AI agents operate business systems, the AI is likewise deciding and acting on its own. Any organisation that has begun handing the execution of work to AI, robots or otherwise, will benefit.

How does this differ from the other AI seminars?

NXGN-020 "The Physical AI Shock" covers the technology and its impact on industry, NXGN-040 "Responsible Management and AI" covers executive responsibility and regulation, and NXGN-060 "Next-Generation Risk Management" covers the organisational mechanisms for preparing against AI risk. This seminar focuses on the operational role: who stops the AI, and how.

Request

Interested in hosting this seminar at your organisation?

We are happy to tailor the schedule, capacity and format (in-person or online) to suit your needs. Enquiries from HR and corporate planning departments are most welcome.

Enquire about this seminar