Webinar · 120 minutes

The Agentic AI Scale

How far do you let the machine work on its own? Most companies use "agentic AI" as a sales label, not as a measured level of autonomy. This session delivers the scale that separates automation from real autonomy, the eight levels that make it up, and the vocabulary the market uses to sell one as the other.

What the session delivers
120minone session
8levels, in 4 classes
4stages of the scale
1two-round test

The full scale, with typical market tools dated at each step — to recognize which level the work already running in your company sits at today.

The problem

"AI agent" has become a label, not a measured level

Gartner estimates that, of every thousand vendors selling themselves as AI agents, fewer than a hundred and thirty actually are — the rest is traditional automation with a conversational layer on top. The fifteen-minute demo does not tell the two apart, and most companies have no yardstick of their own to try.

2/3
what the market has already named

of companies say risk is what stalls scaling

It is not a lack of available technology. The stated obstacle is granting the machine autonomy without knowing how far it can go unsupervised.

<13%
Gartner · agent washing

of "agentic" vendors are actually agentic

The sales vocabulary moved up faster than the technology delivered. Without a scale of its own, the buyer has no way to hold anyone to the difference.

The instrument

Four stages, eight levels, a single yardstick

The scale arranges automation and autonomy on a continuous axis, not as a binary choice. Each stage groups two levels, and each level has a real market tool attached to it — dated, so it does not age as a promise.

Class I

Assistance

The machine suggests; the person decides and executes. It is the most common step today, and the easiest to mistake for "we already have agentic AI".

Class II

Supervised automation

The machine executes a closed task; the person approves beforehand or reviews afterwards. The two-round test separates this level from the next.

Class III

Conditional autonomy

The machine decides and acts within a declared boundary, with no prior approval — and the boundary is what has to be written down.

Class IV

Broad autonomy

The machine redefines its own approach within the objective. It is the step most companies describe having, and few actually operate.

The limits, stated in the session

What this session does not do

A hundred and twenty minutes deliver the public scale. The next layer — how much control each level requires, and where the autonomy ceiling sits for each process — belongs to a more advanced step.

Does not deliver

The control axis and the autonomy ceiling by process

The scale says which level the machine is at. How much control that level requires, and where the ceiling that is not negotiable sits, is content reserved for the masterclass.

Does not promise

That moving up a level is always an improvement

The scale describes capability, not recommendation. Deciding how far up to go is a business decision, and it depends on the process, not on the technology available.

Format

A hundred and twenty minutes, live, with real cases

It works for a group from a single company or for an open audience. Each class on the scale comes with a public case, with its source, so the level can be recognized without relying on the sales pitch of whoever sold the tool.

Length

120 minutes

It fits a slot in an executive calendar, without requiring half a day to be blocked off.

Who it is for

Those who decide, not those who implement

It requires no technical background. It requires that the person has to approve, contract, or defend an agentic AI project in front of someone.

Where it leads

To the masterclass, if it makes sense

The webinar delivers the public scale. The Executive Judgment Masterclass takes on the next decision: how much control each level requires and where the ceiling sits for each process.

What stays with you

One question to take away

What your company calls "agentic AI" today — which of the eight levels is it actually at?

Do you know what level the AI your company already calls agentic is at?

A hundred and twenty minutes, and the eight-level ladder becomes a working instrument, not deck jargon. Tell me whether it is for you or for your team and I will reply with the possible dates.

I want to attend this webinar
Before or after this session