The AI Everyone Used but Underperformed
Thursday 8 October — Art of Leadership (I) · AI in the firm
The table
The case
Subtitle: A fictional leadership case for discussion. Source: The owner's own case (Dražen Kapusta, 2026-09-13), verbatim — the eight questions are his
The founder had approved the subscriptions himself.
He liked new tools. He had built the company by moving early, trusting capable people and recognising opportunities before larger competitors took them seriously. When his employees began experimenting with AI, he encouraged them.
Six months later, almost everyone was using it.
Sales prepared proposals faster. Marketing produced more content. Finance could explain a spreadsheet in seconds. Department heads arrived at meetings with polished presentations. The company had never sounded so articulate.
At the quarterly review, each department had an AI success story.
Then the finance director put the business results on the screen.
Revenue was below plan. Margins had barely moved. Important proposals still waited for approval. Customers still repeated the same information to different departments. The leadership team still carried unresolved decisions from one meeting into the next.
The founder looked around the table.
“Help me understand something. If all of us are working better, where is the improvement in the business?”
Nobody had a convincing answer.
A Faster Version of the Same Company
The first explanation was time.
People needed more practice. The tools were changing quickly. Several managers suggested another training programme. Someone proposed a more advanced model. Marketing wanted an integrated platform. Sales wanted agents.
The founder asked his team to walk him through one recent customer opportunity.
It had started well. A salesperson used AI to research the prospect and prepare a proposal in an afternoon. Previously, that work would have taken two days.
But the proposal assumed a delivery schedule that operations had never approved.
Operations returned it with questions. Sales generated a revised version. Finance discovered that the pricing depended on volumes the customer had not committed to. Another revision followed.
By the time the proposal reached the founder, it was eighteen pages long.
The customer had asked for a clear answer to three questions.
The document was impressive. The opportunity was going cold.
“What exactly did we make faster?” the founder asked.
“The first draft,” the sales director replied.
“And what happened after that?”
“We followed the usual process.”
That was where the conversation changed.
What the Tools Had Never Been Given
Over the next few days, the founder sat with people while they worked.
He found genuine progress.
A customer service colleague had built a useful way to summarise long complaint histories. A junior analyst was exploring scenarios that previously required help from a senior manager. An experienced salesperson used AI to challenge her assumptions before important meetings.
He also found people spending an hour correcting something they could have written themselves in twenty minutes.
One manager generated extensive reports because producing them had become easy. The team receiving those reports now had more to read.
Another used AI to prepare recommendations but supplied very little context. The answers were fluent and broadly sensible. They rarely helped with the difficult choices facing this particular company.
The founder began asking a different question:
“What does the AI need to know to do this job well?”
The answers were scattered.
Customer history sat in emails. Pricing exceptions lived in the finance director’s memory. Operational constraints were understood by three experienced employees. The reasons behind previous strategic decisions had never been recorded.
Even the leadership team disagreed about which customers the company should pursue.
They had given people access to powerful tools while leaving much of the company’s business logic implicit.
Now those tools were producing confident answers around the gaps.
The Founder’s Part in the Problem
There was another uncomfortable discovery.
The founder himself was a bottleneck.
He wanted initiative, but managers knew he might overturn a decision late in the process. They prepared more analysis to anticipate his objections. AI made that preparation easier, so the packs grew longer.
He had asked for speed without clarifying which decisions people could make without him.
He had asked for better commercial judgement without explaining the trade-offs he made instinctively.
He had approved an AI budget without choosing a business result for which anyone would be accountable.
At the next management meeting, he put his own contribution on the table.
“I helped us buy the tools. I have not yet helped us change how we work.”
That admission made the discussion more useful.
The sales director acknowledged that more proposals had not necessarily meant better opportunities. Marketing admitted that publishing more frequently had become a goal of its own. Operations pointed out that nobody had measured the work created downstream by faster output upstream.
The company had evidence of activity. It had pockets of individual productivity. It still needed to establish where those gains improved performance across the whole business.
The Dinner Where the Reports Stayed Closed
That evening, the founder joined a small group from his COTRUGLI generation.
They had learned together, worked through demanding assignments, travelled together and stayed at enough dinners for professional conversations to become honest ones. They knew his ambition. They also knew when he was making a difficult problem sound simpler than it was.
He began with his usual summary.
“AI adoption is strong. We now need to optimise the implementation.”
One of his classmates smiled.
“Tell us what actually happened.”
He told them about the eighteen-page proposal.
A fellow founder recognised the pattern immediately. Her company had accelerated content production before discovering that approval queues had absorbed much of the time saved.
Another alumnus described a more encouraging experience. His team had selected one recurring customer problem, brought the people involved into the same room and redesigned the work together. AI helped with specific steps. The team measured the time until the customer received a usable answer.
A third asked the question that stayed with him:
“If your people save five hours, what are they now able—and expected—to do with those five hours?”
He had no clear answer.
The conversation continued after dinner. They compared mistakes, challenged each other’s assumptions and offered to bring practical examples to a COlab working session.
The founder could ask questions there that he found harder to ask in his own boardroom. The others could challenge him without turning the exchange into a performance.
That trust gave him something another product demonstration could not: a place to examine his own leadership while working on the business.
A Choice for Monday
By Friday, three proposals were on his desk.
The first recommended upgrading the company’s AI tools and expanding training. Several employees were already reaching the limits of their current setup, and stronger technical support could help.
The second proposed selecting one commercially important workflow and redesigning it across departments. Sales, operations and finance would work together on the journey from customer request to an approved, deliverable offer.
The third recommended giving the strongest internal AI users time and authority to help colleagues. Their practical knowledge was valuable, but much of it remained personal and invisible.
Each proposal addressed a real issue. Each required attention from people who already had full workloads.
The founder could fund some combination of them. He could not give every initiative equal priority.
The finance director wanted evidence of business value within ninety days. The sales director wanted to protect time with customers. Operations wanted fewer late surprises. Employees wanted to know whether the time they saved would create room for better work or simply bring more tasks.
Before Monday’s meeting, the founder needed to make several choices.
Which business outcome would come first? Who would own it across departmental boundaries? What would the team stop doing to make room? How would they distinguish a weak tool from an unclear task, missing knowledge or a poorly designed process?
He also needed to decide what authority he would genuinely release.
On the first page of his meeting notes, he wrote:
“In ninety days, what should our customers and our business experience differently because we use AI?”
This time, he intended to begin with that question.
The CDayZ Leadership Challenge
You are the founder’s leadership team.
AI is already part of everyday work. Some employees have achieved meaningful gains. The company’s overall performance has fallen short of expectations, and the reasons are still contested.
Your task is to choose the next ninety-day move.
The decision must connect a business outcome, a specific change in how people work, the role of AI and clear ownership. It must also explain how the company will learn from its strongest practitioners and from the wider COTRUGLI tribe.
The founder needs a decision he can lead on Monday.
Discussion Questions
The owner's "Eight Questions for the Room":
- Where would you begin the diagnosis?
- Which business outcome deserves the first ninety-day commitment?
- What would you change in the customer proposal process?
- What must the founder change personally?
- What should happen to the time people save?
- How would you make the company’s best AI practices transferable?
- How could the COTRUGLI tribe help this team reach a better decision?
- What would convince you to expand, revise or stop the initiative after ninety days?
What evidence would help you distinguish limitations of the AI itself from missing context, weak skills, poor coordination or unclear leadership?
Choose one outcome and explain why it matters more than the other improvements competing for attention.
Where should AI contribute, where is human judgement essential, and which steps should be simplified or removed?
Which decisions, expectations or habits currently depend on him, and what authority would he need to give others for performance to improve?
How would you turn individual efficiency into better customer service, stronger commercial work or greater employee capability—and establish whether that actually happens?
What should experienced employees teach the system and their colleagues, and how would you recognise their contribution while keeping that knowledge useful and current?
What would you bring to a trusted COlab working session, whose experience would you seek, and what practical work could you do together afterwards?
Define the evidence you would require, including how you would check whether gains in one department had created costs or extra work elsewhere.
A Lesson for the Next COTRUGLI Generation
AI can help a person complete a task before a company is ready to benefit from it.
Turning that capability into business performance requires leaders to make priorities explicit, improve the flow of work, share knowledge and assign responsibility for outcomes. It also requires them to examine the habits they bring to the system.
The COTRUGLI tribe offers a setting in which that work can become more honest and more practical. People who have learned, worked, eaten, travelled and celebrated together can share unfinished thinking, reveal mistakes and help each other test a better approach.
The founder had started by asking what AI could do.
He now had a more demanding leadership question:
What would they need to learn, decide and change together to make it perform?
Your comment on this chapter
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.