Method note1 of 2
Confidentiality as architecture: why confidential data never comes in
The three-step perimeter protocol.
AI-Ready · Conselheiro AI-Ready
Oversee AI without executing — and use AI in your preparation.
What AI adds to the board member's craft, practiced where it happens: in simulated board meetings, with AI agent-characters and real pushback. Eleven oversight competencies — without executing.
What AI adds to the director's craft
Fiduciary duty in general, corporate procedure and collegial dynamics are prerequisites for anyone who takes the seat — they are not the subject of the program. What is practiced here is what AI adds to the craft: questioning the thesis, sizing the cost, demanding the gate, recognizing the incident and preparing with AI without depending on it.
A complement, never a confrontation, to the director education programs that already exist: what we offer is the table — the simulated meeting in which oversight becomes observable behavior.
Literacy is not your exposure. Your exposure is the duty that cannot be delegated and the capital you approve without its total cost.
What you will learn
Eleven competencies. Each one opens with a sentence of what you will be able to do.
01How AI works, fails and costs
Understand predictive, generative and agentic systems at the level needed to interrogate a management report — origin and quality of the data, failure modes, what drives cost in production — and recognize when the proposed human oversight actually decides, rather than signing off.
02AI value thesis: where AI produces results and where not to adopt
Challenge the AI proposal that reaches the table as a business matter: calibrate the scrutiny to the size of the investment, separate efficiency gains from changes to the model, ask whether the problem really requires a probabilistic model — and stand behind both a cheap exploratory adoption and a “not here, not now”.
03AI economics: what it costs, what scales with use and how capital is decided
Size the magnitude and nature of the cost — fixed per seat or variable per consumption —, recognize where it arises once the system goes into production and decide between releasing the funds, releasing in stages with conditions for continuing, or stopping.
04AI regulation and compliance in board decisions
Recognize the living regulatory map — data protection and automated decisions, the Brazilian bill in progress, the European regime, sector rules, risk management frameworks — applied to the classification of the organization's own use cases, demanding proportional safeguards and knowing when to call in advisors.
05AI use policy: when it is needed, whether it protects enough and what behavior it produces
Judge the internal rule on AI use at the stage the organization is in: whether it is already needed as a formal instrument, whether it protects enough for this moment and what behavior it tends to produce — including the points that touch people: data in circulation, labeling of generated content, image and voice, monitoring.
06Security, AI-assisted fraud and incident readiness
Identify the AI risks relevant to this business — unreviewed content reaching the customer, fraud through synthetic media, insiders being recruited, leaks through unapproved tools, agents with permission to act — and question the risk map for those gaps, demanding readiness in detection, response and communication.
07The duty to be informed when a decision runs through AI
Recognize what AI changes in the duty to be informed before deciding: the opaque layer between the facts and the decision, the machine recommendation whose origin and basis need to be examined, and the AI incident as a matter for inquiry, record and, where applicable, disclosure to the market or the authority.
08Overseeing implementation: from pilot to scale with measured value
Follow, from the board seat, the crossing from pilot to production: require a starting point, a business metric and the condition of the data before starting; distinguish adoption from realized value; spot the metric whose definition changes between two cycles; demand the decision to continue, expand or shut down.
09Using AI in your own preparation as a director
Prepare with AI to reduce the information asymmetry against material curated by management — reading the meeting materials, independent comparison with public sources, assembling the likely line of questioning — with discipline in confidentiality, verification against the source and a record of what was used.
10Deciding with AI in the room without outsourcing judgment
Recognize what the arrival of AI changes at the moment of deliberation: the tendency to accept the machine's recommendation without asking for the premise, the urgency manufactured by market and vendors, and AI-generated analysis inside the discussion — treated as input to be challenged, never as a vote.
11Reading what lies ahead and saying what it changes in today's decision
Read the AI horizon on a documented basis: separate what is already in production from what has been announced and from what is market expectation — and define the observable event that brings the topic back to the agenda.
The eleven competencies are practiced in simulated board meetings, in Teste Ácido and in Preparar Apresentação. Educational content — not legal advice. The map is revised within weeks when the rules change.
The table before the table
Realistic simulations, designed by someone who did this in person, live, for 23 years. In the room, independent AI agent-characters — each with its own role, information and objectives. You decide under time pressure and pushback, with no audience and no political cost.
The CEO. Wants the decision today. Has a preference and does not state it.
The CFO. Asks for the total cost and the sensitivity — and knows when the number has been rounded.
The vendor. Brings market urgency. Some of it is real.
Legal counsel. Reminds you of what the rules require — and what they do not yet say.
You. Decide. No audience, no political cost, with the pushback the real meeting would have.
Preparation
You receive the case material and the triggers — what the table knows, what each character wants, what is at stake. You prepare as you would for the real meeting: with AI, if you wish.
Live meeting
By voice, in real time, with the agent-characters. They disagree, push, withhold, ask for the number. You lead the table to a decision.
Artifact and debriefing
A recommendation, a vote, a plan — the artifact the meeting required. Then the private debriefing: what you did, what it practiced and what remains as a hypothesis.
Confidential data never comes in: the case is built for the simulation. Confidentiality is not a policy — it is architecture.
Differentiators — after the learning
Teste Ácido
Your proposal or position cross-examined before the real meeting — by an AI designed to challenge, not to agree.
Preparar Apresentação
The information you still need to ask for and the likely Q&A from the Executive Committee or the board, assembled before the meeting.
Confidentiality by design
Perimeter protocol, no automatic integrations, isolation per participant, vendor independence.
Content revised in weeks
The topic map comes in by ingestion, with no retraining. When regulation or the market changes, the program changes the same month.
What you receive at the end of a cycle
The debriefing is private and descriptive. It speaks to what you did when facing the cases — not to who you are.
- An overview of your process: how you used the preparation, which prompts you addressed, how you engaged the characters, how you contributed in the meeting and how you led it to a decision.
- Three to five strengths tied to what each simulation practiced.
- Three to five reflections, presented as hypotheses for real situations — never as prescriptions.
No result is shared with third parties. In cohorts contracted by companies, feedback to the contracting company follows what the contract defines, always at cohort level.
What we don't do
- We do not predict performance.
- We do not pass judgment on the person.
- We do not replace the board's or the executive's decision.
- We do not take in confidential data.
- There is no hidden human layer in the flow: what is AI is declared as AI.
Live board — Boardroom
Sourced data on AI and board oversight.
Sourced data1 of 3
35%
of directors say their boards have already incorporated AI into oversight.
ⓘ Source
PwC, Annual Corporate Directors Survey, Oct 2025 — 556 directors
self-reported data
How to enroll
Individual cycle
For those who decide on their own. Online enrollment, by credit card.
- 3-month cycles
- Access to the platform, the cycle's simulations, Teste Ácido and Preparar Apresentação
- Private debriefing at the end of the cycle
- Renewal in new cycles of 3, 6 or 12 months
- Prices and conditions on the enrollment page
Private cohort
For companies and boards. Ten participants per cohort; direct contract and invoicing — not through the site.
- Cohort of 10 participants per 3-month cycle
- Cases adapted to the sector, with no confidential company data
- Feedback to the contracting company at cohort level, as per contract
- Payment by invoice or boleto (bank slip)
Frequently asked questions
?Do I need to know how to code or understand models?
No. The program is about decisions — formulating, approving, overseeing, defending — and the technical level is that of the decision-maker, not the developer.
?How much time per week?
About two to three hours, between preparation, simulation and the preparation tools. Live meetings last 45 to 60 minutes.
?What is the demonstration?
An individual session in which you experience a segment of a simulation and receive feedback. You request it on the site; scheduling is done by email.
?Can I use my company's content in the simulations?
No — by design. The cases are built for the simulation and no confidential data enters the platform. In private cohorts, the cases are adapted to the sector, not to the company.
Get to know the program before you decide.
Request a demonstration: one individual session, with feedback and no commitment.