The Questions Boards Are Asking About AI
And the ones they aren’t, but should be.
Boards are asking about AI. Often, the questions are accompanied with demands for management to move faster. Two years ago, the question at most board tables was whether AI was a trend worth tracking. Today it has migrated to strategy, risk, and operations. The conversation has arrived, but it is incomplete. Boards are standing at the front of the building, studying the entrance, and drawing conclusions about the whole structure.
The stakes are not abstract. Seventy-eight percent of enterprises have now adopted AI in at least one function — up from 55% in 2023.[1] Yet only 29% report significant ROI from generative AI, and Morgan Stanley found just 21% of S&P 500 companies could cite a measurable business benefit at all.[2] The gap between adoption and outcome is widening. Meta — one of the most aggressive AI-first corporate pivots on record — recently had its CTO describe the company’s AI reorganization as “atrocious.”[3] The headlines are full of equally bad attempts at AI transformation that start and end with “AI can do the work — how many people can we lay off?” More than 275,000 tech workers were cut in AI-cited layoffs in 2024–2025; a Harvard Business Review analysis found most of those cuts were driven by anticipation of automation, not demonstrated gains.[4] As a result: people hate AI. A February 2026 NBC News poll found just 26% of registered voters view it favorably — less popular than Trump, ICE, and both political parties.[5]
Look past the noisy fails, and AI is a capability story, a liability story, and the most consequential infrastructure shift of our time. Boards approaching it only through the cost lens are not being prudent, the specter of rising token costs notwithstanding. They are being unimaginative at precisely the moment that imagination is the competitive advantage.
What I hear when I sit in these rooms — as a board member and in conversations with fellow execs and investors — is a pattern. The questions feel rigorous. They are just zoomed into one corner of a much larger problem.
The Questions Boards Are Actually Asking
To be fair: no one has really figured this out. Directors have lived through a brutal sequence — a pandemic that took valuations to the moon then reimposed gravity, inflation that brought ZIRP-induced tech euphoria to a screeching halt, political upheaval that made long-term planning feel almost quaint. LPs are grumbling. Exits are lengthening. Boards are exhausted and they are cautious.
So it is understandable that when something as vast and fast-moving as AI arrives, many boards look at what everyone else is doing. The herd is culling headcount, focused on efficiency, treating AI as a cost-reduction tool. Following the herd feels safe and defensible. And yet it is a missed opportunity at a scale that does not come around often.
The Headcount Question
“If we automate [business function], can we save $[number] on heads this year?”
So many companies have announced layoffs in the name of AI that it has become a cliché. The logic goes: automate the process, need fewer people, let them go. But where fear of job loss dominates, employees hide their AI use rather than showcase it. The technology that should be a shared growth engine becomes a competitive hedge that cannibalizes morale and stalls adoption. Savings now does not mean savings later. The people most capable of driving AI-enabled innovation are often exactly the ones being let go in its name.
The Adoption Rate Question
“Are our people using AI, and how much?”
This is the right question to start with — getting from non-user to basic user is a real achievement. The problem is treating it as an ending indicator rather than a starting one. I talk to frustrated executives whose organizations rolled out an expensive platform, made everyone log in, and called it done. Adoption rate cannot measure downstream progress, and it invites gaming: employees tokenmaxxing to climb a leaderboard while the underlying work remains unchanged. The better question is what people are using AI for, and whether that use is moving outcomes that matter.
The Velocity and Competitors Question
“Can we go faster? What are our competitors doing?”
Both reasonable, but incomplete. Velocity without direction is expensive urgency, and “our competitors are doing it” without underlying vision is a pressure campaign, not a strategy. Employees can tell the difference between leadership with a genuine vision and leadership reacting to a board conversation. The better questions underneath both of these are: where are we going, and how will we know when we have arrived?
The Questions Boards Are Not Asking
Enterprise AI adoption is holistic change management spanning four dimensions: culture, process, technology, and governance. Boards are good at applying upside pressure — pushing management to move faster, experiment more, not fall behind. What most are not doing is applying equivalent downside discipline. That asymmetry is the core governance problem, and AI makes it dangerous at a speed that traditional governance cycles were not designed for.
Culture
IS OUR EXECUTIVE TEAM ACTUALLY USING AI — OR JUST MANDATING IT FOR EVERYONE ELSE?
Leadership teams that drive AI adoption while personally remaining non-users create two problems: they cannot make informed decisions because they have no felt sense of what AI does, and they signal that AI is something that happens to other people. Does every leader — CEO included — have a personal vision for how AI applies to their domain?
ARE WE BEING INTENTIONAL ABOUT OUR AI CULTURE — OR JUST LETTING IT DRIFT?
AI culture — how an organization collectively relates to AI tools, what norms exist, what the implicit rules are — determines what your AI policy actually means in practice. Culture is what happens when no one is watching. Your AI culture should mesh with your overall culture. If your AI culture is drifting, your strategy and governance is drifting with it.
Process
ARE OUR AI TOOLS CHANGING HOW WORK GETS DONE — OR SITTING ON TOP OF BROKEN PROCESSES?
AI applied to a broken process creates a process that can break faster into more pieces. Boards should be asking whether management has mapped underlying workflows, identified where human judgment is genuinely required, and designed AI in rather than bolted it on. This is where most AI initiatives quietly fail. While the board doesn’t need to get into the weeds, management should be able to articulate an overall process overhaul thesis that is more thought out than “we tell each team to figure it out.”
ARE WE DESIGNING FOR HUMAN JUDGMENT — OR JUST ASSUMING IT WILL HAPPEN?
As AI takes on more of the execution layer, the consequential question is not what AI is doing — it is what humans are still responsible for, and whether those people have the context, time, and authority to exercise real judgment when it matters. Human oversight tends to be assumed rather than designed. “Someone will catch it” stops being a governance answer as AI scales. The board should be asking who owns this decision, how it is made consistently, and how it holds as use expands.
Technology
WHERE IS OUR BUILD VS. BUY STRATEGY — AND WHO OWNS IT?
Most organizations are making these decisions tool by tool, team by team, without a governing framework. Both instincts — buy without thinking, or sit it out until the market settles — make the same mistake: they delegate the thinking to someone else. Every vendor designs for their average customer. Far from that average, and the defaults are quietly wrong for your organization — and you may not know it until an output goes out under your name you would not have approved. The board question is not build or buy. It is whether management has enough of a build capability to make an informed buy decision, and whether they are actively ceding decisions about what the organization actually is to a vendor’s defaults.
ARE WE TAKING ANY BIG SWINGS — OR JUST INCREMENTAL NUDGES?
Incrementalism is the path of least resistance and, increasingly, the path of competitive irrelevance. The board should be asking not just what AI tools are being deployed, but what the organization is doing that it could not have done two years ago. If the answer is a list of efficiency improvements, the organization may be optimizing the existing model rather than building the next one. Each organization only has bandwidth for two or three such initiatives at a time. The board should be asking which ones they are — and whether management is actually swinging.
Governance
WHEN SOMETHING GOES WRONG, DO WE KNOW HOW TO RESPOND — AND WHO IS RESPONSIBLE?
Not if. When. AI systems fail confidently — they hallucinate, encode bias, degrade without warning. In early deployments I’ve worked on, we treat two principles as non-negotiable: human oversight at every consequential decision point, and failure that surfaces loudly rather than silently. Most organizations have neither. “Someone will catch it” stops being a governance answer the moment AI scales. Who owns the response? What is the escalation path? Has the board discussed what a material AI failure looks like for your specific organization? These are the questions regulators and plaintiffs will ask. Getting ahead of them is critical in 2026.
IS OUR GOVERNANCE KEEPING PACE WITH THE BUSINESS — AND WHO OWNS THAT?
Governance that was designed for last year’s AI stack is not governing this year’s. The question is not whether a framework exists — most organizations have one, at least on paper. The question is whether it is keeping pace, has real resources behind it, and whether someone in leadership owns it with genuine accountability. If nobody can say what the framework covers now that it did not cover six months ago, and what mechanisms we have in place to ensure it stays current, your governance is already stale.
The Best Question — And A Structural Fix
How are we getting smart on AI — and who on our board owns that accountability?
This is perhaps the most important question of all — and one not enough boards are asking about themselves. The regulatory environment is accelerating and will not slow down. Liability is hardening across jurisdictions. Opting out because “technology is not my thing” is no longer defensible. The board does not need to be technically expert — but it needs enough literacy to ask the right questions, recognize when an answer is incomplete, and know when to bring in outside expertise.
Governance has a reputation problem — the boring cousin of strategy, added to the board deck as an afterthought. That has to change. In the AI era, governance is not the handbrake; it is the accelerator. The structural answer is an AI committee with standing comparable to audit — not a working group, not a standing item bolted onto an already full agenda. A dedicated mandate covering the stack, the governance model, the risk surface, and management’s approach across all four dimensions, with access to independent technical advisors and the authority to require changes, not just recommend them.
Boards are already grappling with something genuinely unprecedented — technology moving faster than any governance framework was designed to handle. The directors who build their own literacy, ask the uncomfortable questions, and push for structure before the crisis rather than after it are the ones who will look back on this moment as the one where they led.
Natalie Kim is the founder of Inflection Advisory and works with organizations on AI acceleration and governance. As VP Legal at Omnidian she led a full-arc enterprise AI adoption.
The Judgment Layer publishes on AI governance, board accountability, legal intelligence, and the judgment no algorithm is taking from you.
[1]Writer, Enterprise AI Adoption in 2026: Why 79% Face Challenges Despite High Investment (2026). Available at: writer.com/blog/enterprise-ai-adoption-2026/
[2]Morgan Stanley, AI Is Now a Macro Variable. Are You Positioned? (2026). Available at: morganstanley.com/insights/articles/ai-market-trends-institute-2026
[3]Wired (Lauren Goode and Paresh Dave), Meta CTO Andrew Bosworth Admits the Company’s AI Reorg Was ‘Atrocious’ (2026). Available at: wired.com/story/andrew-bosworth-meta-employees-unrest/
[4]Harvard Business Review, Companies Are Laying Off Workers Because of AI’s Potential, Not Its Performance (January 2026). Available at: hbr.org/2026/01/companies-are-laying-off-workers-because-of-ais-potential-not-its-performance
[5]NBC News, Poll: Majority of Voters Say Risks of AI Outweigh Benefits (February 2026). Available at: nbcnews.com/politics/politics-news/poll-majority-voters-say-risks-ai-outweigh-benefits-rcna262196