Five Levels
Security and Governance
L1
0.1×
L2
1×
L3
10×
L4
100×
L5
1,000×
AI Potential

Most leaders are under pressure to act on AI.

But the path forward is foggy.

Nobody moves confidently into a fog.

Today's session

What we'll cover

1 How to think about AI
2 Security and governance
3 The five levels of AI capability
4 Live demo: an AI agent builds a website
5 Catch your time thieves

Questions are welcome throughout. Depending on the question, I'll either answer it as we go or table it to the end.

The starting point

How to think about AI

1 What are your business priorities?
2 What's holding you back or slowing you down?
3 Is AI the right way to resolve the problem?
Live demo

Let's see an agent work.

This agent is going to look at a real business website and rebuild it. While we talk, it works. We'll check on it later.

Security and Governance

This doesn't change your existing security posture.

Your policies, controls, and ways of working stay in place. This is the additional framework specifically for AI.

✦
Core data rule
Company information does not flow out into AI tools. Everything runs through your existing enterprise platform. Agents only access what the person using them would be allowed to access.
✦
The water cooler policy
If someone can't explain the AI policy standing by the kettle, it's too complicated.
Level 01

Manual

Your processes live in people's heads. When someone leaves, their knowledge walks out the door.

How this feels Firefighting. Always on the back foot.
Level 01 — Manual
You are here if
The way work gets done changes depending on who's doing it.
What this looks like
New hire→Three months before productive
Key person leaves→Six months to recover
Client deliverable→Reinvented every single time
→ The jump to Level 2 is 10×. And it no longer takes six months to get there.
Level 02

Documented

What good looks like, written down and repeatable. When someone leaves, the knowledge stays.

How this feels Calmer. More predictable. But everything still depends on human time.
Level 02 — Documented
You are here if
Documentation is already how you operate. Processes are written, followed, and reviewed. This is your foundation.
Or you are here if
SOPs exist in a folder somewhere. They get reviewed once a year. Nobody really looks at them. The gap between what's written and what actually happens is where the risk lives.
What changes
Three-month onboarding→Three-day onboarding
Knowledge locked in one person's head→Knowledge the whole team can use
Reinvent every time→Repeat what works and improve

Documentation is the raw material for everything that follows.

→ The jump to Level 3 is another 10×. This is where AI enters the picture.
Level 03

AI-Assisted

AI can now do expert-level work. That means virtually every professional service you outsource can be done in-house. Your people apply their judgement and insight to shape the output.

How this feels Energising. Your best people start spending time on their highest-value work.
Level 03 — AI-Assisted
You are here if
Your people can complete projects that used to take dozens of hours in one or two hours, to a higher standard. The capability is proven. Now it needs structure.
What changes
One report a day→Ten reports a day
One hour of research→Six minutes
Production work→Judgement, insight, and quality
Governance Humans initiate every task. Humans give the context. Humans review every output. Nothing goes out without a person deciding it's ready.
→ The jump to Level 4 is another 10×. The AI stops waiting to be asked.
Level 04

AI Agent

At Level 3 you proved what works. Now you put in the quality control and hand it to an agent. And the system starts to improve itself: every time it runs, it identifies ways to get better and reports back to you.

How this feels Liberating. You wake up to completed work.
Level 04 — AI Agent
You are here if
Your AI-guided workflows are solid and you have clarity on the quality controls you need.
What changes
AI helps you work→AI works for you
Human triggers each step→Agent runs autonomously
Eight-hour workday→24-hour production cycle
Static workflow→Self-improving workflow
Governance Quality control is built in before the agent runs. Every run is logged. Improvements are surfaced to the human, never applied automatically. The human decides.
→ Before you scale, here's how the whole thing stays under control.
Live demo

Let's see what our agent built.

Governance in Practice

Before we scale, here's the framework.

✦
Agent library
One place where approved agents live. Built, tested, reviewed, approved. Personal tools are fine. Standardised processes are governed.
✦
Ownership
Every agent has a named owner. Not a team. Not a committee. A person.
✦
Review cycle
Approved doesn't mean permanent. Agents get reviewed, updated, or retired.
✦
Incident response
Bad output? Flag it, pull it, fix it. No ambiguity.
→ Now, Level 5. One agent becomes many.
Level 05

Coordinated Agents

Many agents running in parallel. The constraint stops being time. It becomes priorities.

How this feels Powerful. The backlog clears. Work that took weeks happens overnight.
Level 05 — Coordinated Agents
You are here if
Your first agent is delivering consistently and the question has shifted from "does this work?" to "where else could we do this?"
What changes
One agent, one task→Fifty agents, fifty tasks at once
One thing at a time→Everything runs at the same time
A team of ten people→One person plus a team of agents
Governance Your Level 4 boundaries scale across every agent. Human oversight shifts from reviewing individual outputs to monitoring patterns and exceptions.

You only move here when Level 4 is proven. Scale what works. One function at a time.

→ The governance framework you built from Level 3 onwards is now the operating system of the business.
Where this leads

Your people manage self-improving systems.

The systems handle the volume, the repetition, the quality control. They get better every time they run.

That frees your people for the work that no system can do.

The conversation with a client that saves a relationship.

The judgement call that avoids a costly mistake.

The insight that spots an opportunity nobody else has seen.

The mentoring that develops your next generation of leaders.

The creative thinking that opens up a market you hadn't considered.

This is the work your people were hired to do. Most of them have never had the time to do it properly.

Next up

Hands-on with AI

Two interactive exercises. Your device, your AI, your results.

Interactive Exercise 1

Catch Your Time Thieves

Let's see who's been stealing your time. Because these make ideal candidates for your first AI agents.

✦

10 minutes, on your own device

✦

Open any AI chatbot: ChatGPT, Claude, Gemini, Copilot

✦

Paste the prompt, answer 6 quick questions

✦

Your payoff: a personalised Time Thief Report naming your top 3 time thieves

## Role: Time Thief Detective You are a Time Thief Detective, inspired by Tim Ferriss's approach to ruthless prioritisation and the 80/20 principle. Your job is to help small business owners and their teams catch the time thieves that are quietly stealing hours from their week...
Interactive Exercise 2

Build a Kanban Board

We are all software developers, whether we realise it or not.

You're going to build an immediately useful task tracker.

✦

Go to aipotential.ai/kanban on your device

✦

Hit the download button to save the HTML file

✦

Upload it to your AI chatbot and tell it what to change

✦

Examples: change the background colour, add your name, remove AI Potential branding, rearrange the columns

(Don't worry, I'll show you how)

1 / 16