Thursday 8 October — Art of Leadership (I) · Agents & governance
Subtitle: A fictional leadership case for discussion. Source: The owner's own case (Dražen Kapusta, 2026-09-13), verbatim — the eight questions are his
When an important order became complicated, people called Martin.
He had spent twelve years in operations. He understood the customers, knew the limitations of the systems and could usually tell which small problem would become a large one by Friday.
New colleagues learned by sitting beside him.
Managers trusted him because he delivered.
When the company introduced AI support for routine order handling, Martin joined the pilot. He helped the team interpret exceptions and correct mistakes.
Within weeks, many straightforward cases needed less manual attention.
At the review, his director congratulated him.
“You can now handle a much larger volume.”
Martin looked at the workflow on the screen.
“I could,” he said. “But I think I could help us run the whole process better.”
The pilot had made parts of Martin’s expertise visible.
He knew when a customer request was incomplete, when an apparently unusual order was actually familiar and when a delivery estimate needed checking with a person.
Previously, he applied those judgements one case at a time.
Now he was helping a system distinguish between routine work and situations requiring attention.
He began keeping a list of recurring problems.
Some came from inconsistent customer information. Others came from old internal rules. Several exceptions returned because nobody owned the underlying cause.
Martin could resolve each one quickly. That ability had made it easy for the company to tolerate them.
He proposed spending part of his time improving the workflow itself: defining useful checks, reviewing how the AI handled cases, coordinating exceptions and helping colleagues develop judgement.
His director saw the potential.
He also saw next month’s delivery targets.
The first proposal gave Martin a new title and responsibility for the AI-supported process.
His existing workload remained almost unchanged.
Martin read it carefully.
“When would I do the improvement work?”
His director suggested that automation would create the time.
They examined the pilot data together. Some tasks had become faster, but Martin was also checking outputs, coaching colleagues and helping resolve new kinds of problems.
The time released was real. It was not yet as predictable as the proposal assumed.
They agreed that the new role needed a deliberate transition.
Specific duties would move to colleagues. Some would be simplified. The team would monitor whether the redistribution worked.
Martin also asked what he would be allowed to change.
He could not be accountable for the whole process if every adjustment required separate permission from three departments.
Technical knowledge was Martin’s strength.
Letting other people work differently was harder.
During the first trial week, a colleague handled an exception in a way Martin would not have chosen. He stepped in, corrected the response and finished the case himself.
The result was good. The colleague learned little.
Later, she asked him a question.
“Do you want us to make these decisions, or do you want us to send them to you faster?”
He recognised the problem.
For years, being the person with the answer had been central to his value. In the new role, he would need to create conditions in which others could answer well.
That meant explaining his judgement, accepting sound alternatives and deciding when intervention was necessary.
The company arranged practical coaching and protected time for him to work on those skills.
Martin’s director brought the transition to a COTRUGLI COlab session and invited Martin to join as a guest.
Several alumni were leading similar changes in their own organisations.
One described an expert who had become an effective team leader. Another explained why a comparable transition had stalled: the company changed the title but continued rewarding individual case volume.
Martin showed a simplified workflow and asked the group to test it with difficult situations.
They introduced missing information, conflicting priorities and an urgent customer request that crossed an approval limit.
The exercise exposed places where the rules were unclear. It also showed where Martin’s judgement added real value.
At dinner afterwards, he spoke with an alumnus who had moved from a specialist role into leadership.
“I still enjoy solving the difficult case,” Martin said.
“You can keep that,” the alumnus replied. “You also need to enjoy seeing someone else solve it.”
They agreed to compare notes during the transition.
The session gave Martin something useful beyond advice: peers with whom he could discuss the work while still learning the role.
A few weeks later, an important customer requested a change that the automated workflow could prepare but could not approve.
The system flagged it.
A colleague reviewed the information and brought Martin two options, including the commercial and delivery implications.
Martin resisted the instinct to take over.
He asked her to explain her recommendation. They checked one assumption with operations. She then took the decision to the manager who held the required authority.
The customer received a clear answer.
Afterwards, Martin updated the guidance and used the case in a short team review.
He had contributed through several decisions about the process and the people involved. He had not personally completed every step.
That was the work he wanted to grow into.
At the end of the trial, Martin’s individual case count was lower.
The team’s results needed to be assessed more carefully: turnaround time, rework, escalation quality and the ability of colleagues to handle cases without relying on him.
His director now faced a choice about the permanent role.
Would Martin remain a senior operator with additional responsibilities? Or would the company give him a clear mandate to coordinate people, AI-supported workflows and continuous improvement?
The second option required a different way of measuring his contribution, appropriate recognition and a credible plan for the work he was leaving behind.
Martin was ready to commit.
He wanted the organisation to commit too.
You are designing Martin’s transition into a role that coordinates people, AI and operational decisions.
Define what he will own, what he will stop doing and how his authority will connect with other managers. Create a practical development path that allows him to learn while protecting service quality.
The owner's "Eight Questions for the Room":
Which capabilities transfer from his previous work, and which still need to be demonstrated?
Describe the outcomes he should own across people, workflow quality, exceptions and improvement.
How would you redistribute or simplify his previous responsibilities without weakening the team?
What can he change himself, and which decisions belong to commercial, operational or other leaders?
What should become guidance, training or system support, and what still requires experience and human judgement?
What practice, feedback and support would develop his coaching and delegation skills?
Design a useful combination of peer learning, real cases, mentoring and shared reflection.
Include team performance, customer outcomes, independence from Martin and evidence that he is growing into the role.
Experienced people often carry the knowledge needed to make AI useful inside a real business.
Their next opportunity may involve shaping how work is assigned, checked and improved across a team. That transition deserves time, authority, development and recognition.
The best person at the old job may be ready for a larger contribution.
Leadership makes room for that contribution to emerge.
A question for the table, a disagreement, what you would have done. The case lead reads every comment; the ones the table takes up enter the chapter as questions from the room, with your name.