AI for Leaders: a Method for Better Decisions

Artificial Intelligence has already entered organisations. For a leader, the question is no longer "should I use it?" but how can I use AI to prepare a decision without delegating my judgement?

This was the foundation of Artificial Intelligence Applied to Leadership and Strategic Management, taught at Escola de Negócios – ENB. It was not a technical course, but a way to learn how AI can support analysis, preparation and communication.

What can AI do in a decision?

AI can:

• organise scattered information;
• summarise positions;
• identify unanswered questions;
• propose scenarios;
• challenge assumptions;
• compare alternatives against criteria;
• prepare a communication.

It cannot take responsibility for the leader, fully understand the organisation's culture or predict the future. A well-written answer can also be wrong.

A six-stage method

1. Frame the decision

Write the decision as one concrete sentence. "Improve sales" is a broad goal; "decide whether to open a new sales channel next quarter" is a decision.

Identify the deadline, owner and people affected.

2. Separate facts, assumptions and unknowns

Create three lists:

• what is known and can be verified;
• what is assumed;
• what still needs to be discovered.

Ask AI to organise the information without mixing the categories. This separation reduces the false appearance of certainty.

3. Define criteria

Before discussing alternatives, select criteria such as cost, impact, time, risk, reversibility and team capacity. Specify which ones carry more weight.

4. Explore scenarios

Use AI to build a baseline, favourable and adverse scenario. For each, ask for signals that would indicate which one is emerging.

The goal is not to guess. It is to prepare responses before pressure increases.

5. Run a critical review

Ask for arguments against the preferred option, second-order consequences and the perspectives of different stakeholders. Then verify everything that depends on external facts.

6. Record and communicate

Document the decision, reasons, assumptions, risks, owner and review point. AI can prepare versions for the board, team or customers, but the final message must reflect the leader's intent.

Example prompt for exploring a decision

Act as a decision facilitator, not a decision-maker. Based only on the information below, organise facts, assumptions and unknowns. Compare the three alternatives by cost, impact, time, risk and reversibility. Identify the information most likely to change the decision. Do not select an option.

This instruction keeps AI in a supporting role. It also requires explicit criteria instead of asking for a generic recommendation.

Where AI should not decide on its own

Take particular care with decisions involving:

• employment or assessment of people;
• personal data or confidential information;
• legal, financial or health consequences;
• criteria that may reproduce discrimination;
• actions that are hard to reverse;
• incomplete or disputed information.

In these cases, use specialists, review mechanisms and documentation proportionate to the risk.

To structure adoption before reaching sensitive decisions, see the method for applying AI at work.

How to bring AI into leadership routines

Begin with moments of preparation:

• agendas and questions for a meeting;
• summaries of reports;
• proposal comparisons;
• risk maps;
• communication drafts;
• retrospectives on a past decision.

These cases build fluency without surrendering operational control. As the team gains experience, it can create approved templates and a library of good examples.

Leaders and organisations can assess their starting point with the SuperHumano Academy AI Maturity Index. SuperHumano AI provides programmes on AI for Leadership and Strategic Management and executive consulting to connect technology, priorities and governance.

Frequently asked questions

Can AI recommend a decision?

It can produce a recommendation, but that does not make it accountable or correct. Use it to structure alternatives and questions; the decision still requires context, validation and human responsibility.

How can I avoid over-trusting an answer?

Separate facts from inferences, require sources when appropriate, ask for opposing arguments and verify relevant information outside the tool.

Which data may a leader share?

Only data permitted by the organisation's rules and the tool's terms. Remove identifiers and confidential information wherever possible.

Context note: the original edition of this course began on 25 February 2026, live online, with 12 hours of training and DGERT certification through ENB. That edition has ended; the method remains applicable.