The Real Reason Your Boss Wants AI: Plausible Deniability, Not Better Decisions
Organizations adopt AI in high-stakes decisions partly because it creates accountability ambiguity—a buffer between decision-makers and responsibility that humans cannot provide.
Corporate AI adoption is driven not only by efficiency gains but also by a structural desire to create accountability buffers. The EY Canada case—where hallucinated citations passed through normal review—illustrates how AI tools can generate plausible but flawed advice that is difficult to attribute or blame. This accountability gap represents asymmetric risk that organizations may be knowingly accepting. Effective AI procurement requires the same due diligence applied to human consultants, including credential verification and liability requirements.
When GPTZero published its investigation in 2024, the finding was specific and verifiable: EY Canada had published a cybersecurity report in which most citations were hallucinated—fabricated references that appeared authoritative but did not correspond to any real source. The report had passed through normal corporate review processes and appeared on a major professional services firm's public platform. What makes this incident worth more than a footnote is what it reveals about the structural position AI now occupies in corporate decision-making. The standard assumption is that organizations adopt AI tools because they improve decisions—better analysis, faster processing, cost reduction. That assumption is not wrong, but it is incomplete. A separate dynamic is at work in high-stakes contexts: AI provides something organizations have always quietly wanted when risk is elevated, which is a buffer between decision-makers and accountability. Professional service firms apply credential verification to human consultants. A cybersecurity expert advising on strategy carries professional liability, regulatory oversight, and reputational stakes that create personal consequence for errors. AI systems carry none of these. When an AI tool generates advice that proves flawed, the question of who is responsible becomes genuinely ambiguous—and ambiguity, in organizational politics, is a form of protection. This is not a claim about executive motives based on individual psychology. It is a structural observation: the accountability gap created by AI adoption in high-stakes decisions is real, and it has economic consequences. Organizations that move quickly to deploy AI after a failure or near-miss may be responding to the incentive structure rather than the underlying problem. Speed of adoption signals action to stakeholders; accuracy takes longer to verify and is harder to demonstrate. The EY Canada case is an extreme example—hallucinated citations are a clear failure mode—but the pattern extends further. AI-generated advice that is plausible but wrong in ways that are harder to detect than invented references represents a larger and less visible risk. The citations were caught because they were verifiable. Advice that is internally consistent but based on flawed assumptions may pass through the same review processes without detection. What this means for procurement decisions is straightforward: the same due diligence applied to human consultants—credential checks, reference verification, professional liability requirements—should apply to AI tools being used in consequential decisions. Without it, organizations are accepting asymmetric risk exposure. They gain the appearance of rigor while losing the accountability infrastructure that makes rigor meaningful. The adoption decision itself should also be examined. If your organization's AI procurement accelerated following a failure or crisis, it is worth asking whether the driver was operational improvement or the distribution of risk. Both can justify the same action. Only one makes the organization more capable.