OnBoard: 78% fail AI audits

AI governance is turning into the real test of whether corporate AI spending means anything. A lot of companies have approved budgets, launched pilots, and told investors they are moving fast. But the hard part is not buying models. It is proving who owns the decisions, what controls exist, and what happens when the system gets something wrong. (grantthornton.com) That is the gap Grant Thornton put numbers on in its April 13, 2026 survey release — and the numbers are ugly. ### What actually broke here? The headline number is simple: 78% of executives said they do not have strong confidence they could pass an independent AI governance audit within 90 days. (grantthornton.com) This was not a tiny pulse check either. Grant Thornton surveyed 950 C-suite and senior business leaders across 10 industries between February 13 and March 18, 2026. (grantthornton.com) ### What is an “AI governance audit” in practice? Basically, it is the question every board should already be asking — can the company explain how an AI system works well enough to defend it? That means showing how decisions are made, who is accountable for outcomes, what controls are in place, and what evidence exists if regulators, auditors, customers, or courts start asking questions. (grantthornton.com) Grant Thornton’s framing for this is the “AI proof gap” — companies are scaling AI they cannot clearly explain, measure, or defend. ### Why are so many failing the smell test? Because management knows governance is the problem, but many companies still are not treating it as the priority. In the survey, 46% said governance and compliance failures are a leading cause of AI underperformance. But only 11% said risk and compliance is the function needing the most focus to hit their AI ambitions. (grantthornton.com) That is the tell — leaders can see the leak, but they are still funding speed over plumbing. ### Where is the ownership gap? It sits across the C-suite. Grant Thornton’s follow-up report on May 8 said CIOs and CTOs, COOs, CFOs, and CEOs are often operating from different realities. One striking example: 39% of CIOs and CTOs said their workforce is fully ready for AI adoption, but only 7% of COOs agreed. (grantthornton.com) So the people installing the technology and the people running day-to-day operations are not even describing the same organization. ### Why does that matter to boards? Because AI risk is not just a tech issue anymore. If one team buys the tools, another team absorbs the workflow disruption, and nobody owns the evidence trail, the board is left with exposure and no clean line of accountability. Grant Thornton says that is how companies end up scaling AI “without anyone accountable for what it produces.” That is not a compliance footnote — it is a governance failure at the board level. (grantthornton.com) ### Is this only about avoiding disaster? No — and that is the part boards should care about most. The survey says organizations with fully integrated AI are nearly four times more likely to report revenue growth than companies still piloting, 58% versus 15%. So governance is not just brakes. It is what lets a company actually press the gas without flying blind. (grantthornton.com) ### What should directors take from this? The practical takeaway is boring but important: treat AI more like financial reporting and less like an innovation side project. Boards need named owners, documented controls, regular review cycles, and evidence that management can produce on demand. If a company cannot explain an AI decision path to its own board, it probably cannot explain it to an auditor either. That is the whole story. (grantthornton.com) ### Bottom line? The news is not that companies are using AI. Everybody knows that. The news is that many companies still cannot prove they are governing it, even as they keep spending more on it. Turns out the winners may not be the fastest adopters — they may be the firms that can show their work. (grantthornton.com 1) (grantthornton.com 2) (grantthornton.com 3)