Why AI Claims Are Now a Board-Level Test of Judgment
95 percent of directors expect AI to affect their business within a year—so why is it still absent from most board agendas? That is not a technology question. It is a judgment question, and for boards, the gap is now hard to defend. According to NACD, nearly all directors see AI as relevant soon, yet only 28 percent say it is a regular feature in board conversations (NACD, 2023).
That mismatch creates a governance vacuum. In a quarterly review, a management team can present “AI-enabled” growth, “AI-driven” efficiency, or “AI-powered” customer experience, and the language alone can push a board toward approval. Capital gets committed. Timelines get shortened. Risk assumptions get softened. When those claims are thin, the cost is not abstract: boards can misdirect investment, miss control failures, and signal to investors that oversight is weaker than the strategy deck suggests. This article gives directors a practical due diligence lens for separating real AI transformation from AI washing before those errors compound.
The Risk Is Not Hype Alone
AI washing is often treated as a messaging problem. It is more serious than that.
When executives overstate what AI is doing in the business, the distortion spreads quickly. Strategy discussions start from false premises. Budget choices favor theater over capability. Risk oversight gets framed around aspiration instead of operating reality. Investor trust is especially fragile here, because markets can tolerate experimentation far more easily than they tolerate claims that outrun evidence.
This is why AI now belongs inside the broader discipline of board governance, not as a side conversation for the most technical director in the room. The board does not need to build models, evaluate code, or debate architectures. It does need to ask whether the claimed AI capability is real, whether it is material to performance, and whether it is governed well enough to justify the risk taken in its name.
Boards Need Evidence, Not Fluency
A common mistake is to assume the answer is more technical education. Some of that helps. But fluency is not the same as oversight.
The board’s role is narrower and more demanding: require proof. What process changed? What decision improved? What control was added? What result can be traced to something more than a relabeled analytics project or a pilot that never left the lab? Those are the questions that expose whether management is describing a business capability or borrowing the language of AI to create momentum.
The pressure on directors is only going to increase. If almost every board expects AI to matter, the real issue is no longer awareness. It is discernment. And that raises the next question: what, exactly, should directors treat as AI washing—and what should they not?
What Does AI Washing Actually Mean for Directors?
The claim-versus-execution test is the right starting framework here because most organizations now describe a wide range of digital improvements as AI. The evidence shows the label is often doing more work than the underlying capability, with CFA Institute Research and Policy Center defining AI washing as the misrepresentation or exaggeration of AI use in products, services, or operations (CFA Institute Research and Policy Center, 2025).
That matters for directors because the board is rarely judging code. It is judging whether management is presenting a real operating capability or a story attached to familiar software.
A Board-Level Definition That Actually Helps
In practice, AI washing usually looks ordinary. A rules-based workflow becomes “intelligent automation.” A dashboard with predictive scoring becomes “AI-driven decisioning.” A software vendor adds a generative feature to an existing product, and suddenly the company’s entire AI strategy is framed as transformed.
None of those moves is automatically deceptive. Some are useful. The board problem begins when the language implies a step-change in how the business works, while the evidence shows only an added feature, a pilot, or a repackaged analytics layer.
AI washing often involves claims that overstate the presence, sophistication, or impact of AI relative to what is actually deployed (CFA Institute Research and Policy Center, 2025)
That is the distinction directors need to hold onto: AI-enabled activity versus AI-transformed operations. The first adds capability at the edge. The second changes process design, decision rights, staffing, controls, customer experience, or unit economics in a measurable way.
Where Directors Usually Get Misled
Take a quarterly review at a mid-market manufacturing company. The COO reports that the service function is now “AI-powered” because technicians receive automated maintenance recommendations. The board hears modernization. But if the recommendations are rarely used, if scheduling still runs manually, and if downtime has not moved, then the business has not been transformed. It has acquired a feature.
This is why terminology is a weak test. Substance is stronger.
For directors, the practical question is simple: what changed because of the claimed AI capability that would not have changed with standard automation, better reporting, or a conventional software upgrade? If management cannot answer that cleanly, the label is ahead of the execution.
A credible answer usually shows operating consequences. Faster cycle times. Fewer exceptions. Better pricing decisions. Lower claims leakage. New revenue from a product that could not previously exist. Without that chain, “AI” is just a narrative wrapper.
And this is where oversight gets harder than it sounds. If the label is easy to apply and the operating proof is slow to surface, how does a board avoid mistaking ambition for evidence—or skepticism for discipline?
Why the Hype Gap Makes Oversight Harder Than It Looks
72% of organizations now say they use AI in at least one business function—so if nearly everyone has an AI story, what exactly is a board supposed to treat as proof (McKinsey, 2024)? When AI use becomes this common, the label stops being a differentiator. It becomes background noise.
That is what makes oversight harder than many directors expect. The challenge is no longer spotting companies that talk about AI. It is judging whether the talk reflects a real operating shift or simply the market’s new default language.
Common Claims, Thin Signals
The pressure starts with prevalence. If most organizations report some level of AI use, management teams can present AI adoption as table stakes rather than a claim that still needs examination. A board packet says “AI embedded in customer service,” “AI in forecasting,” or “AI in internal productivity,” and each phrase sounds plausible because, statistically, it is plausible.
79% of organizations expect AI to transform their organization within three years (Deloitte, 2024)
That expectation matters more than it first appears. It means boards are often hearing transformation narratives long before the underlying economics, controls, and workflow changes have had time to mature. Ambition arrives early. Evidence usually does not.
In a budget review at a regional healthcare provider, for example, a CFO may ask for accelerated spending on a “generative AI transformation” in patient communications. The board hears urgency, competitive risk, and possible efficiency gains. But six months later, the actual deployment may still be limited to draft message generation with heavy human review, no redesign of staffing, and no measurable change in response times. The claim was not necessarily false. It was simply ahead of the facts.
Urgency Distorts Judgment
That timing gap gets worse because 68% of executives say generative AI is a top strategic priority over the next 12–18 months (PwC, 2024). Priority language changes boardroom behavior. It compresses diligence. It makes ordinary questions sound slow, even obstructive.
This is where boards get trapped. If directors push for proof, they risk being framed as behind the curve. If they accept strategic language at face value, they risk approving spend, partnerships, and public claims on immature foundations.
The real oversight problem, then, is not hype alone. It is sequencing—narrative first, proof later. And when management says AI is already strategic, what should the board ask to see: a roadmap, or operating evidence?
What Evidence Should a Board Ask Management to Show?
The Five-Dimension Evidence Test is the most useful boardroom filter I know for AI claims. Without it, directors end up judging confidence, vocabulary, and slide design—while the real questions about substance, ownership, and impact go unasked.
Picture a quarterly review at a regional retail company. The CEO says AI is already improving merchandising and customer retention, and the board is asked to approve more spend before the holiday cycle. That is the moment to slow the room down and ask for evidence across five dimensions: claim clarity, technical substance, operational materiality, governance evidence, and business impact.
Start With Claim Clarity and Technical Substance
First, make management state the claim plainly. Not “AI is transforming forecasting,” but: which use case, in which process, for which decision?
That sounds basic. It is not. CFA Institute Research and Policy Center notes that AI washing often shows up when firms blur what AI actually does, or imply sophistication that is not present (CFA Institute Research and Policy Center, 2025). A credible narrative should therefore name four things without hedging: the specific use case, the data feeding it, the model or workflow involved, and the executive who owns the outcome.
Ask in plain English:
- What decision is this system helping make?
- What data does it rely on?
- Is this a model, a rules engine, or a human workflow with AI support?
- Who is accountable if it underperforms?
If management cannot answer those questions cleanly, the board is not looking at a capability. It is looking at a label.
Then Test Whether It Is Embedded or Merely Announced
The next issue is whether the AI is actually inside operations. Boards should ask to see process evidence, not just presentations, vendor demos, or marketing language.
That means workflow maps, adoption rates, exception logs, retraining or monitoring routines, and proof that frontline teams are using the output in live decisions. If the claimed system improves service quality, show where it sits in the service process. If it supports managers through continuous feedback, show how that guidance enters routines, reviews, or staffing choices. NACD is explicit that boards should focus on how AI affects strategy, risk, and oversight—not just the technology itself (NACD, 2023).
The board’s job is not to verify that AI exists. It is to verify that it is governed, used, and tied to a real business process.
Finish With Governance Evidence and Business Impact
Governance evidence is where weak claims often fail. Ask for model risk controls, escalation paths, human review points, and reporting lines. Who signs off on changes? Who monitors drift, bias, or failure? What gets reported upward, and how often?
Then ask the hardest question: what moved in the business because of this? Not activity. Results.
Show the cycle-time change, margin effect, loss reduction, conversion lift, or retention improvement. Show the baseline. Show the period measured. Show whether the gain held after launch. If management cannot separate operating impact from experimentation, the board may be funding promise rather than performance.
And even when the evidence looks tidy, another problem usually appears first: the small inconsistencies, the vague ownership, the metrics that never quite connect. Are those normal growing pains—or the earliest signs of AI washing?
Which Red Flags Usually Reveal AI Washing First?
13% of U.S. employees use AI daily. For boards, that should immediately sharpen one question: if day-to-day use is still this uneven, how much of the “AI transformation” story is operating reality and how much is announcement (Gallup, 2026)?
You have seen the moment. In a quarterly review, a VP at a regional financial services firm clicks to a slide titled “AI Strategy Acceleration,” and the room fills with confident phrases—AI-enabled workflows, intelligent decisioning, next-generation automation—while no one can quite tell what system is actually live.
Red Flag 1: The Language Gets Abstract Fast
The first warning sign is vague language. Not because executives should speak like engineers, but because management should be able to explain, plainly, what is being called AI, where it sits in the workflow, and what human decision it changes.
When that explanation never arrives, the label is doing the work. A board hears “AI in underwriting” and later learns it is a rules engine with a new interface. Or “AI in service operations” turns out to mean draft responses that staff rarely use. The problem is not semantic purity. It is that unclear language makes weak claims hard to challenge and easy to approve.
Red Flag 2: “Strategic” but Ownerless
The second red flag is more serious: AI is described as strategic, yet no one names the owner, the control point, or the escalation path.
That is not a documentation gap. It is a governance gap. If a model fails, produces poor recommendations, or creates customer harm, who decides whether it is paused, retrained, or withdrawn? Boards should hear a crisp answer. Research from McKinsey shows that 47% of respondents say their organizations have experienced at least one negative consequence from using gen AI (McKinsey, 2024). Negative consequences without clear ownership become board problems very quickly.
This is where risk oversight stops being abstract. If management cannot show who is accountable when the system goes wrong, the strategy is ahead of the controls.
Red Flag 3: The Case Lives in the Future
The third red flag is a business case built almost entirely on future promise. The presentation is full of pipeline value, strategic positioning, and long-term upside, but current outcomes remain unmeasured.
That usually means one of two things: the initiative is still early—which is fine—or the company is using ambition to cover for weak adoption and uncertain impact. Boards need to know which. Is this a real capability with emerging evidence, or a story with a budget line? That distinction matters even more when AI is said to be material to the business—because once materiality is claimed, the next question is unavoidable: material where, exactly, and by what proof?
How Do Boards Test Whether AI Is Material to the Business?
The Materiality Triangle starts with a hard fact: 85% of employers plan to adopt new technologies, including AI, by 2027 (World Economic Forum, 2023). Without a test for materiality, boards end up treating broad adoption as proof of strategic importance—and that is how capital gets spread across initiatives that sound current but do not change enterprise outcomes.
The right question is not, do we use AI? It is narrower and tougher: does AI change results that matter to this business?
Impact, Ownership, Measurability
A claim becomes material when it affects one of five things in a way the board can see: revenue, cost, risk, customer experience, or the operating model. If management says AI matters, ask where the effect shows up in the P&L, the control environment, the customer journey, or the way work is actually done.
That sounds obvious. It is not.
In a budget cycle at a mid-market retail company, a chief digital officer may argue that AI-driven personalization is now “core” to growth. The board should not debate model design. It should ask three plain questions. Did conversion improve? Did customer retention move? Did merchandising or marketing decisions change because teams now work differently—not just faster?
50% of employers expect AI to increase their business’s overall growth rate over the next five years (World Economic Forum, 2023)
That expectation explains the pressure. It does not settle the case. Growth potential is not the same as current materiality.
Core Engine or Layered Feature?
This is where many boards get misled. A company can have real AI use and still have non-material AI.
If the technology sits at the edge of an existing process—drafting copy, summarizing calls, suggesting next actions—it may be useful without being central to value creation. Material AI is different. It changes how the business wins, serves, prices, underwrites, manufactures, or allocates labor. It becomes part of the engine, not an accessory.
That is also where AI accountability matters. If AI is truly material, ownership cannot be diffuse. Someone should own the outcome, the controls, and the decision to scale, pause, or redesign.
Boards should therefore test every major AI claim against the same triangle: impact, ownership, measurability. If one side is missing, the initiative may still be promising—but it is not yet material.
And once a board decides AI is material, the standard changes. Is oversight still occasional, or does governance now need to operate with the discipline of any other enterprise-critical capability?
What Does Disciplined AI Oversight Look Like in Practice?
Boards lose money when they approve AI stories that never become operating reality. They also lose trust—inside the company and outside it—when employees see grand claims, weak follow-through, and no one held accountable.
If AI washing is a claim problem, the board’s answer is not technical theater. It is discipline.
Make AI Oversight Part of Governance, Not a Side Conversation
In practice, disciplined oversight starts with a simple shift: stop treating AI as a special topic that appears only in strategy sessions. Treat it as a recurring board governance issue that touches capital allocation, risk review, operating performance, and management credibility. That is consistent with how NACD frames the board’s role: connect AI to strategy, oversight, and enterprise risk rather than isolating it as a narrow technology matter (NACD, 2023).
That changes the boardroom posture. The question is no longer, Are we doing something with AI? It becomes, What claim is management making, what evidence supports it, who owns it, and what result should we expect to see by the next review?
A regional services company offers a familiar example. During a budget cycle, the CEO backs an “AI-enabled productivity” program and asks for faster approval. A disciplined board does not argue about model design. It asks which workflows will change, which executive owns adoption, what control failures could emerge, and what operating result would show the effort is real rather than branded software spend.
Revisit Claims Until They Either Hold Up or Fall Apart
The most useful board habit is follow-through. Not one hard meeting. A sequence.
CFA Institute Research and Policy Center makes the core risk plain: AI washing thrives when claims outpace what is actually deployed or achieved (CFA Institute Research and Policy Center, 2025). Boards counter that by revisiting the same claims over time and comparing promise with evidence. What was presented as transformational in spring should show up later in process adoption, control reporting, customer outcomes, or financial performance. If it does not, the board should say so directly.
The test of oversight is not whether management can describe AI confidently. It is whether the board can track a claim from announcement to operating proof.
This is where many boards either strengthen management discipline or quietly weaken it. When directors ask again—same claim, same owner, same expected outcome—they signal that rhetoric has a shelf life.
Boards do not need to be technical experts. They need a repeatable evidence standard, the patience to apply it, and the nerve to keep asking when the story sounds better than the results. In your next review, will AI be discussed as a narrative—or examined as a governed business capability?
Frequently Asked Questions
What is AI washing and why is it a concern for boards?
AI washing is the exaggeration or misrepresentation of AI use in business operations, where claims overstate the presence, sophistication, or impact of AI. It is a concern for boards because it can lead to misdirected investments, weak oversight, and false confidence in AI-driven strategies without real operational change.
How can boards distinguish real AI transformation from AI washing?
Boards can distinguish real AI transformation by demanding clear evidence of operational changes such as improved processes, decision-making, or measurable business impact, rather than accepting vague or marketing language. They should verify that AI is embedded in workflows, governed properly, and producing tangible results beyond pilot projects or added features.
What types of evidence should boards require to validate AI claims?
Boards should require evidence across five dimensions: claim clarity, technical substance, operational materiality, governance controls, and measurable business impact. This includes clear use cases, data and model details, integration into operations, risk management practices, and proof of improved outcomes like faster cycle times or increased revenue.
Why is AI oversight a judgment issue rather than just a technical one?
AI oversight is a judgment issue because boards do not need to evaluate technical details like code but must assess whether AI claims reflect real business capabilities and risks. Effective oversight focuses on verifying evidence of AI’s material impact and governance rather than relying solely on technical fluency or buzzwords.
What are common red flags indicating AI washing during board reviews?
Common red flags include vague or abstract language about AI capabilities, lack of clear operational changes, missing accountability for AI outcomes, absence of governance controls, and no measurable business improvements. These signs suggest that AI claims may be more narrative than substantive.
