AI in the boardroom: the right questions before adopting it

The AI conversation has reached the boardroom. The question is no longer "should we use it?" but "how do we use it without exposing ourselves?". Before adopting any AI tool to support board decisions, three questions demand clear answers.

1. Where do the answers come from?

A generic language model answers with what it learned from the internet. That's useful for drafting an email, but unacceptable for deciding on a debt covenant. The minimum bar for a board is traceability: every claim must be traceable to a specific passage in a company document — with the document, page and section cited.

In Directa, this rule is architecture, not a promise: no figure leaves the system unless it appears literally in the company's documents or follows from them by arithmetic that is shown. If the source doesn't say it, the system states that explicitly. The absence of information is itself a signal the director should know about.

2. Who sees my data?

Board information is among the most sensitive a company holds: unpublished results, strategy, conflicts, compensation. Before adopting AI, the board must know precisely where data is processed, whether it's used to train third-party models (the answer must be no), and what role-based access controls exist.

For companies with strict regulatory or confidentiality requirements, the on-premise option — where the entire system runs inside the company's own infrastructure — eliminates the question at the root: the data never leaves the perimeter.

3. What happens when the model doesn't know?

The most underrated risk of generative AI isn't that it gets things wrong, but that it gets them wrong with confidence. A board-grade system must be calibrated: report its level of certainty, prefer a good question over a bad claim, and earn urgency with evidence, not adjectives.

A calibrated question to the board is better than an incorrect claim delivered with a confident tone.

The right yardstick

AI for boards shouldn't be judged by chatbot standards but by the standards of a good senior analyst: does it cite its sources? Does it distinguish what it knows from what it infers? Does it protect confidentiality? Does it prioritize what matters? When those four answers are yes, AI stops being a risk to manage and becomes what it should be: an advantage for better governance.