For decades, the composition of a board of directors was resolved with a relatively stable combination of profiles: one financial, one legal, one industry. That combination reasonably covered the risks that a board had to monitor. Today, it leaves one of the most relevant — and those that are approved most frequently without proper scrutiny: whether a data or artificial intelligence project has a solid foundation, or whether it is a well-presented recommendation built on data that no one has validated.
The traditional counselor profile does not cover this risk.
A financial profile can assess whether the projected return from an AI project is reasonable compared to the capital requested. A legal profile can assess regulatory and contractual risks. An industry profile can assess whether the project meets a real business need. None of the three, by design, is capable of assessing whether the data on which the recommendation was built is reliable, or whether the proposed model effectively answers the business question it is meant to solve.
This is not a criticism of those profiles — it is a description of an structural void. The advice was designed to monitor the risks that dominated the corporate agenda for decades: financial, legal, and market-related. Data and AI are a relatively new category of risk in that agenda, and most of the advice has not yet incorporated the profile that it deserves.
What happens when that profile is not on the table
Without an individual capable of questioning the technical underpinnings of a proposal, the typical reaction of a board to a data or AI project is not to reject it out of caution. It is to approve it based on how convincing the presentation was—because there is no one in the room with the judgment to question it in any other way.
That can take just minutes. A well-crafted business case, an attractive return projection, and a recommendation presented with confidence are enough to get the green light, even when no one has verified whether the source data are reliable or whether the success metric has been clearly defined. The result is not necessarily a bad project — it is a project approved without the oversight that committed capital should require.
Technical criteria do not mean that the entire council becomes technical.
The most common mistake when identifying this gap is to assume that the solution requires all members of the board to acquire knowledge of data science or artificial intelligence. That is not the case, and demanding it would be unrealistic for most boards.
What is needed is different: at least one person at the table with the sufficient judgment to ask a limited set of questions that reveal whether the underlying work was done rigorously. Not technical questions about model architecture or algorithms — business questions applied to a data or AI project: what decision is going to change, how the success will be measured, and how reliable the source data is.
That profile could be an independent advisor with prior experience in digital transformation, the chair of an expanded audit committee, or any member willing to develop that approach deliberately. What matters is not the title held, but that there is someone with the explicit responsibility to ask those questions in every relevant proposal.
What that profile should be able to do, in concrete terms
Defining the profile as «someone with technical expertise» is too abstract to be actionable. In concrete terms, the person responsible for this role within the board should be able to do at least three things in response to any data or AI proposal.
First, distinguish between a return projection built on a validated pilot and one built on optimistic assumptions without evidence. The difference is not always visible in the executive summary that reaches the board, and you need to ask directly how each figure was obtained.
Second, identifying when a recommendation depends on data that is likely not governed—multiple sources without a common definition of customer or product, historical information without quality assurance, or time series with gaps that no one mentioned. There is no need to audit the data line by line to ask this question; there is a need to know that it needs to be done.
Thirdly, verify that a defined success metric exists before approval, not after implementation. A project that is approved without anyone being able to name how success will be measured almost always ends up being evaluated retrospectively and subjectively, which makes it practically impossible for the board to learn anything useful from that decision for the next one.
The cost of not having this profile is not immediate.
One of the reasons why this gap persists for so long in a board is that its cost does not manifest itself immediately. A project approved without the appropriate technical judgment does not necessarily fail visibly in the short term—often it progresses, consumes budget, and produces ambiguous results that no one can clearly attribute to a good or bad original decision.
That kind of ambiguous outcome is, paradoxically, more costly in the long run than an outright failure. An outright failure generates a course correction. An ambiguous outcome over time erodes the board’s confidence in its own ability to approve these types of projects—which often results in one of two equally problematic reactions: approving them increasingly with less scrutiny, out of habit, or halting investment in data and AI entirely, out of excessive caution, without either of these resolving the original vacuum.
How to identify if the council has this gap
There is a simple test: in the last data or AI proposal that the board approved, did anyone ask how reliable the source data was before approving the capital? If the answer is no, or if no one remembers asking that question, the gap exists, regardless of how solid the other profiles represented at the table are.
Another sign is the speed of adoption. A data or AI project that is approved in the same session in which it is presented, without follow-up questions about the quality of the data or the success metric, generally did not undergo scrutiny—it was accepted based on the confidence it generated from the person who presented it. None of these signals require technical knowledge to identify; they simply require someone on the board to pause and review the minutes of the session with that question in mind.
Frequently asked questions
Is it necessary to hire a consultant with a data science profile? Not necessarily. It is more effective to develop the criteria in an existing member of the board with experience in digital transformation or technological projects, rather than to hire someone external without knowledge of the organization’s specific business.
How many data or AI proposals does a council need to see before this gap becomes relevant? One is enough. The risk of approving a project without the appropriate criteria does not depend on how many similar projects come after — it depends on the fact that the capital committed in that single approval is backed by a verified recommendation.
What happens if the board decides not to include this profile? The board remains in place, but there is a lack of oversight over a category of risk that will grow, not decrease, as more business decisions depend on data projects or AI.
Is this gap exclusive to formal management advice? No. The same pattern appears in board committees, investment committees, and any collegial body that has to approve capital for data or AI projects without an explicit mandate of technical oversight. The name of the body changes; the structural void remains the same.
Next step
At MaIA by Scanda, we work with board members and management committees that need to build the internal technical criteria to evaluate data projects or AI, and with the reliable database — Data Quality and Data Lake — that those criteria need to operate. It’s not about replacing the board’s judgment with a permanent external consultancy; it’s about leaving the ability to ask these questions autonomously in every future proposal in place. If your board is about to approve a project of this kind, let’s talk.
Español 
English