By Dr. Salvatore Scuderi
A leadership team receives an AI-generated market assessment. The argument is clear, the language is confident, and the recommendation appears sensible. Before anyone acts, however, someone needs to ask what evidence supports the conclusion, which assumptions shaped it, and what the system may have missed.
That moment is where AI leadership becomes practical. Access to an AI tool does not settle how much authority its output deserves. Leaders still have to decide what to question, what to verify, and who is responsible for the outcome.
My work in management and my doctoral research on AI-based Hybrid Intelligence in strategic decision-making have led me to focus on this relationship between technological capability and human judgement. For organisations considering executive education, it is a useful place to begin: what must leaders learn to make sound decisions when AI becomes part of their work and their teams?
Start with a decision worth improving
An organisation can begin with a specific recurring decision: which customer segment to prioritise, which operational problem to investigate, or which investment proposal deserves further review – and which KPIs should guide it. The task is to understand how that decision is made today before introducing AI into it.
Who provides the evidence? Where do delays occur? Which assumptions are disputed? What would a better outcome look like? These questions give an AI initiative a business purpose and make it possible to assess whether the additional technology helps.
Consider a management team reviewing an expansion proposal. AI could organise market information, generate alternative scenarios, or identify inconsistencies in the business case. The team would still need to assess the reliability of the information, the relevance of local context, and the consequences of committing resources. This is an illustrative example, not a client case.
Make the division of work explicit
Hybrid Intelligence describes the combination of human and artificial intelligence in a shared task. In practice, the quality of that combination depends on how the work is divided and how each contribution is checked.
A useful design starts by naming the tasks that AI may perform and the decisions that require human approval. A system might draft an analysis while a designated manager checks its factual basis. A team might use AI to develop options while retaining responsibility for selecting and defending a course of action.
The phrase human in the loop is too vague on its own. A reviewer needs enough time, knowledge, access to evidence, and authority to reject an output. If those conditions are absent, human approval can become a formality.
Treat confidence as something to examine
A fluent answer can be persuasive even when its reasoning is incomplete. Leaders need habits that help them examine an output without accepting it simply because it sounds convincing.
One habit is to separate facts from assumptions and recommendations. Another is to ask which evidence would change the conclusion. A third is to examine a credible alternative explanation. These are familiar disciplines of management, but they need deliberate attention when analysis can be produced quickly and in large volumes.
AI can also be used to challenge an initial position. A leadership team can ask it to identify weaknesses in a proposal or develop a competing scenario. That exercise is useful only if the team checks the response and remains willing to question its own preferred answer.
Build trust around the task
Trust in AI should be specific to the use case. A system that helps prepare a meeting summary does not automatically deserve the same reliance in an investment recommendation. The information available, the consequences of error, and the ability to verify an output all matter.
For a selected use case, leaders can define what acceptable performance means, how errors will be recorded, and when a person must intervene. Confidence can then be revised as experience accumulates. A successful demonstration is a starting point for evaluation; it cannot establish reliability across every business situation.
Connect governance to everyday work
Governance becomes useful when employees can apply it to an actual task. They need to know which information they may enter, which tools are approved, when outputs require review, and where to raise concerns.
Managers also need a clear owner for the business process and a route for escalation. Responsibility can become difficult to trace when a recommendation passes through several systems and people. Agreeing these arrangements before expanding a use case makes later decisions easier to examine.
Executive education can provide a setting in which leaders work through these questions together. A shared business case allows participants from operations, HR, IT, and senior management to compare their assumptions and agree how a process should work.
Make learning visible in the workplace
An AI leadership programme should give participants work they can apply. Each participant could bring a real decision process, identify a suitable AI contribution, and define the review needed before acting on an output.
The organisation can then assess the proposed change against a baseline. Measures might include preparation time, factual errors, rework, or the quality of the evidence used in a decision. Attendance and participant satisfaction provide useful feedback about a course, but they cannot establish business impact on their own.
A practical first step is to select one decision process, bring its owner and relevant colleagues together, and examine where AI could help. That conversation provides a concrete basis for deciding what the leadership team needs to learn next.
Discuss your organisation’s learning needs
If your organisation is exploring AI leadership development, I invite you to discuss a decision or workflow your team wants to improve. We can use that example to clarify the learning needs and explore a suitable executive education approach with EIM.
Book a conversation with Dr. Salvatore Scuderi through the EIM page.