83% of S&P 500 companies now name AI as a risk factor in their filings, up from 12% in 2023. Only 2.7% of their directors have any AI expertise at all.
So almost every board has registered the risk, and almost no board is equipped to oversee it. We have been here before. This is what cyber looked like between 2015 and 2017: everyone knew it mattered, nobody knew what to ask, and then came the incidents that taught everyone the hard way.
I spent 25 years on the management side of that table, presenting to boards. A board does not need to understand the technology in order to oversee it. It needs to ask the right questions and recognize a weak answer when it hears one. Here are the five I would ask.
Question 1: What business problem does this solve, and what is it worth in shekels
If management brings an AI initiative and cannot name a specific process and a specific number, it is a reputation project, not a business project. A good answer sounds like this: "Invoice processing takes us four days and costs X, and industry benchmarks report reductions of 70% to 90% in back-office processes." A weak answer sounds like this: "Everyone is moving into this and we do not want to be left behind." One more lens on the same question: research from MIT (the NANDA project, 2025) found that most AI budgets go to marketing and sales, while the highest measured return sits in operations and finance. Ask management where your own initiatives fall on that axis.
Question 2: Who owns it in management - and what happens to their KPI
The State Comptroller report of June 2026 found that 68% of AI projects in Israel are stuck in pilot. The leading reason: no owner. A project that belongs to "innovation" belongs to nobody. The board should hear the name of one executive whose personal targets contain the project's metric. If the answer is a committee, there is no owner.
Question 3: What data does this run on, and who is accountable for its governance
The same Comptroller report found that 41% of organizations operate with no data governance at all. That is no longer only an operational question, it is personal exposure. Amendment 13 to the Privacy Protection Law, in force since August 2025, carries fines of up to NIS 3 million and personal liability for officers. In industry surveys, privacy and data governance are already the number 1 barrier to AI adoption at 61.3%, ahead of model accuracy and ahead of integration. The board needs three facts: what data goes into the system, whether it leaves the organization, and who signed off on that.
Question 4: What is the committed success metric, and when is the decision made
This is the question that kills eternal projects in one stroke. Before a single shekel is approved: what is the baseline today, what is the numeric target, what is the decision date, and what happens on that date - scale up or shut down. A project without that clause is not an investment, it is a bet. And from the building side, a vendor who will not commit to a number is telling you something useful about their own confidence in the product. Equally important is who does the measuring. Measurement has to run against a baseline set before go-live, by someone with no stake in the result: not the vendor, and not the team that championed the project.
Question 5: What happens when the system gets it wrong
Because it will. The question is never whether, only what happens next. There are good engineering answers to this one, and I build them with my own hands in the systems I run: double filtering in content moderation, a human in the loop on sensitive decisions, automatic stop rules. The board does not need to follow the implementation. It needs to confirm that somebody designed for the failure. Under Israel's new financial regulation of December 2025, this is no longer voluntary: the supervised entity carries 100% of the responsibility for its AI decisions.
What to do with the answers
None of these five questions requires a technology director, and none requires a Python course. They require exactly what a board is supposed to be doing anyway: demanding ownership, numbers, risk management and decision points. Ask them, and within a single meeting you will know whether the organization has an AI plan or a drawer full of pilots.
You can make it formal, too. Among S&P 500 companies, the share with a board committee responsible for AI went from 6% to 20% in a single year. Not every organization in Israel needs a dedicated committee, but every organization needs the subject on the agenda as a standing item, quarter after quarter, the same five questions against the same numbers. Consistency at board level does more for an organization than any innovation deck ever will.
Worldwide, meanwhile, 76% of large organizations have already appointed a chief AI officer, according to IBM, up from 26% the year before. The gap between the companies that govern this at executive level and those that leave it to "a team that will look into it" keeps widening, and it will show up in the results.
Want to put these five questions to your own AI initiatives, with someone who has sat on the management side of the table and also builds these systems himself? Send me a note, and I will bring a demo along.
