Why AI Investments Risk ‘Performative Governance’ Without Leadership Accountability

Despite massive investments in cutting-edge AI tools, data failures are escalating. Even as enterprises across industries layer on expensive monitoring and compliance software, the crisis persists. It's not a tech problem, it's a leadership crisis. And it's shrouded in what one expert called "performative governance." Now, executives must truly own the outcomes of their data and AI strategy if they want to avoid wasting budgets and eroding trust.

Nachiket Mehta, Head of Data and Analytics Engineering at Wayfair, is a seasoned technology and AI executive. Known for driving data-enabled business transformations, Mehta has reportedly delivered over $480 million in operational cost savings during his career. Mehta said that the industry’s approach to governance is fundamentally flawed, a problem now magnified by generative AI.

• Tools don't govern, people do: For Mehta, the path forward begins with an often-ignored truth: "Tools don't govern, people do." Here, leadership plays an especially important role. Otherwise, the "obvious principle" of accountability is routinely sidestepped, often with severe consequences. For instance, while organizations invest in multimillion-dollar AI tools, few actually succeed, he said. "Responsibility for tool failures is rarely claimed, though such failures occur more often than leaders admit."

• The accountability vacuum: Treating governance as a separate, technical problem, creates an accountability vacuum, Mehta explained. "When asked about governance ownership, the common response is 'Not me, it's that other team.' This indicates a problem. All leaders and employees must be accountable for governance, not just a select few." Unfortunately, finger-pointing culture often prevents true ownership, he explained.

To avoid being seen as the "compliance police," some centralized governance bodies diffuse responsibility until they're essentially powerless, Mehta explained. He described experiencing this firsthand at Wayfair, where the central data team often felt like a "big wall in the middle," getting all the blame when something went wrong but none of the credit when it didn't.

• The federated solution: To break the cycle of blame and inaction, Mehta spearheaded the adoption of a Data Mesh framework at Wayfair. In a departure from centralized control, he embedded data engineers directly into business domains to act as coaches, rather than service providers. "I opted to teach teams how to manage their data, rather than doing it for them. This type of federated model pushes ownership to the source, making data-creating teams directly responsible for quality and governance."

According to Mehta, a culture of proactive accountability is the antidote to "performative governance," or the act of buying tools and writing policies that create the appearance of control without changing underlying behavior. As generative AI floods enterprises with even more unstructured data, they could also create flawed feedback loops that undermine AI initiatives, he said.

• Escaping the project trap: To make governance truly effective, it must be inseparable from business value, Mehta advised. "Governance must be integrated into the same OKRs that drive business value. No need to create separate internal tech transformation projects." Eliminating standalone data and governance projects could also encourage leaders to integrate accountability directly into business objectives, he said.

• Naming the owners: But if central committees and separate projects aren't the answer, then who is accountable? For Mehta, it's the existing business leaders already responsible for outcomes. "Domain leaders, such as a VP of Supply Chain or an HR leader, should own their entire workflow, including data governance."

The biggest challenge for C-suite executives will be disciplined restraint, Mehta concluded. "Leaders must resist the urge to solve technology problems first. Instead, they should begin with an honest business assessment, using a maturity model to diagnose the organization's true state, from basic monitoring to full AI-driven transformation." While often overlooked, this foundational work is what truly separates those struggling at level one from the AI-first, truly transformative digital platforms of the future, he said. "It is the only viable path to the ultimate vision of efficiency: the 'one-person unicorn,' an organization so automated and data-mature, it achieves massive scale with a minimal human footprint."