Who decides, who executes, and who answers for it? When an AI agent moves up a level inside a team, it does not only carry out more tasks — it changes the work of everyone around it. This session delivers the three questions every manager has to be able to answer about their own team, drawn from a real case told in two parts.
The case shows a system that works well for ninety days — and that is exactly where the problem appears, not in an obvious failure.
When an agent moves up a level of autonomy, it does not only execute more — it changes the work of everyone around it. The manager starts receiving a larger share of the hard work, and loses the practice of doing the rest. The effect has been known since 1983 (Bainbridge, "Ironies of Automation"), and it keeps repeating with every new layer of automation.
Two hundred B2B accounts, a renewal-risk system that advances level by level — up to the scene in which a wrong rule, not an obvious failure, multiplies one problem across four consecutive weeks.
The team reorganized itself around the agent without anyone formally deciding it. An important account is lost to signals no system was picking up. The dashboard was green the whole time.
They do not depend on knowing the Renata case to work. They are the three questions that separate a team that has genuinely changed from a team that simply got a new tool.
If the agent acts within a boundary, someone has to review that boundary periodically — or it ages in silence.
When the agent gets it wrong or stalls, the work does not disappear — it goes back to a person. Knowing to whom keeps the exception from becoming a crisis.
The question the Renata case answers in the worst possible way: nobody noticed, because the dashboard measured what the agent did well, not what it failed to see.
What stopped passing through people? Does a human layer come back into the review? Does the next level of autonomy solve the problem or hide it better?
One hundred and twenty minutes deliver the leadership instrument. The technical autonomy scale behind the case — and how much control each level demands — belongs to another step.
The case mentions that the agent "moves up a level", but the eight-level scale that describes this in detail is the content of the other webinar, "The Agentic AI Scale".
The session delivers the right questions for your team, not a universal number. The answer changes by process and by risk.
It works for a group from a single company or for an open audience. The three decisions from part 2 of the case are put to the room before the real outcome is revealed — that is the exercise that changes the conversation afterwards.
It fits a board calendar slot, without requiring half a day to be blocked out.
No technical background required. It does require that the person answers for a team where machine and people already share the work.
The webinar delivers the three leadership questions. The Executive Judgment Masterclass takes on the technical decision that follows: how much control each level of autonomy demands.
On your team, if an agent got something wrong today and no one noticed for three weeks, would the dashboard you look at tell you?
One hundred and twenty minutes, and the three leadership questions become part of the management routine, not an isolated exercise. Tell me whether it is for you or for your team and I will come back with possible dates.
I want to attend this webinar