AI and board preparation — an ongoing inquiry

AI is already in the boardroom.The question is not whether,but which form will prevail — and under what governance.

The debate addresses the AI a company adopts and must oversee. There is a second dimension, still barely codified: the AI a director uses to prepare for the board. That is the one we work on.

The problem

Directors already use AI to prepare. Largely without oversight.

01

A powerful tool, ungoverned

Under pressure, the director falls back on general-purpose AI tools that treat every question the same way. And uploads confidential board packs to consumer services: opaque terms of service on confidentiality, interactions that are traceable and potentially relevant in litigation. A risk that engages the company's interests and the director's individual fiduciary duty at once.

02

Pressure that is real

Board packs grow longer, preparation time compresses. An international survey of 468 directors finds that half report not spending enough time on strategic preparation between meetings.

03

A concrete cost, not theoretical

Generic questions, born of an unstructured prompt, that lower the discussion instead of raising it. The risks of ungoverned AI use in the boardroom were systematized by the Harvard Law School Forum on Corporate Governance in November 2025.

The question is not whether AI enters the boardroom: it already is. It is which form will prevail, and under what governance.

The distinction

Two ways to understand AI in the boardroom.

In the regulatory debate, board AI is treated almost exclusively as an object of oversight. Yet there is a second dimension, where professional discipline is still to be built.

AI as an object of oversight

The systems a company adopts, subject to assessment and compliance with the European regulatory framework. It is the right frame for that plane: the discipline is taking shape, and the board's task is to oversee its responsible use.

AI as the director's tool

The AI a single member uses to prepare for the board. It operates within the individual professional perimeter, across multiple mandates, and engages two planes of responsibility at once: the company's and the director's own fiduciary duty. The discipline is still barely codified. This is the domain we work on.

The two perspectives are complementary. A better-prepared director is in a position to better oversee the company's AI.

The working hypothesis

A grammar, not a classifier.

The system does not answer every question the same way: it differentiates its treatment by the nature of the request. Four modes.

At every stage, judgment remains the director's. The tool does not replace it: it arms it.

Triage

When the answer is already in the documents.

A precise pointer to the material: citation, no elaboration.

Italian Corporate Governance Code 2020, Recommendation 12

Accountability

When what was promised is what matters.

Reconstructs the commitments management made and checks their execution against the current state.

EBA/GL/2021/05 · Bank of Italy, Circular 285

Socratic stress

When the question is still raw.

Structured sub-questions that bring the interpretive question to maturity before the table.

Italian Corporate Governance Code, Principle IX

Amplification

When the question is already mature.

Enriches it with precedents, sector peers and regulatory frameworks.

Catani, 2023

In practice

Socratic stress, at work.

The most demonstrable mode — and the only one tested in the pilot. An example, on the annual approval of an incentive plan.

The director's raw impression

"This incentive plan strikes me as too generous."

At the board table

"…isn't it a bit generous?"

vague, easily dismissed.

The tool gives neither the answer nor the question: it leads the director to find them. The judgment stays theirs.

Not just theory

The grammar is already at work.

One of the four modes — Socratic stress — is already implemented in a working tool, in use on real mandates. It is what makes it possible to run the pilot from day one. The other modes are at varying stages of maturity: defining the grammar in full is part of the work underway.

Want to see it at work? Write to us.
The study underway

Bringing it into focus, on evidence.

Before consolidating the tool, we are conducting a structured study of the risks and opportunities of AI in board preparation. The aim is to understand — on evidence, not intuition — where AI creates real value for board work and where it introduces risk to be governed.

Direct listening

An ongoing programme of interviews with those who live the role: directors, statutory auditors, board secretaries.

Analysis

The academic and regulatory literature.

Observation

Developments in governance practice, Italian and international.

A pilot tests a single mode — Socratic stress — against the anchor of a Harvard study (Kestin, 2025): step-by-step guidance outperforms passive reading. The analogy to board preparation remains a hypothesis to be validated, not an established result.

There are, to date, no Italian empirical studies on the effectiveness of structured preparation tools for directors. That is precisely why the study exists.

Kestin et al., 2025 (Scientific Reports) · Deloitte Global Boardroom, 2024 · Harvard Law School Forum on Corporate Governance, 2025 · Italian Corporate Governance Code 2020 · EBA, 2021 · Bank of Italy, Circular 285

Take part

Take part in the building.

We are building a network of interlocutors — professionals, companies, academic institutions and professional associations — to engage with before and during the design.

Those who take part at this stage help define the approach and the tool, more than they receive it.

Take part in the study

A structured conversation on how you prepare, and — if you wish — a trial of the Socratic mode on a synthetic case. No confidential document enters the system.

Request the reference document

The extended analysis: the distinction, the full grammar, the regulatory anchors in full text.

If the subject concerns you, let's talk.

Frequently asked questions

What we are asked.