The 28% Problem: Why AI ROI Measurement Is Now a Series C Governance Issue

The Headline the Board Saw

A Gartner study published April 7 found that only 28% of AI infrastructure projects fully succeed and deliver return on investment. One in five fail outright. Fifty-seven percent of IT and operations managers have experienced at least one AI project failure.

That number is traveling fast. I have seen it show up in investor prep materials, enterprise procurement documents, and board pre-reads in the past week alone. The question it generates is not technical. It is organizational: which category does this company fall into, and how do we know?

At Series C, that question has a deadline. Enterprise customers are adding AI governance evidence to RFPs. Investors expect measured outcomes, not deployment counts. Getting ahead of this is not optional for the next board cycle. It is a readiness issue.

What It Actually Means

The Gartner data comes from a survey of 782 IT infrastructure and operations managers. The lead researcher, Gartner research director Melanie Freeze, explained the pattern clearly.

The failure data matters. But the measurement gap is the more interesting problem. An analysis published April 8 in AI Magicx found that 86% of enterprises increased their AI budgets in 2025, while only 29% of executives say they can reliably measure the return on those investments.

That gap, where most companies spend more while few measure reliably, is where the governance vulnerability lives. And for Series C companies specifically, it is becoming the thing that separates a clean board narrative from an uncomfortable one.

The finding that matters most for measurement readiness: organizations that establish pre-deployment baselines are 4.2 times more likely to demonstrate ROI later. Not because baselines are magic. Because they force teams to define what “working” looks like before the first deployment decision. That discipline is what generates the answer when a board member or enterprise procurement lead asks for evidence.

The companies I see handling this well are not the ones with the most AI deployed. They are the ones who asked the measurement question first.

Three Questions the Board Will Ask

Here are the three questions I expect to surface in Series C board meetings this quarter. The honest answer to each is probably “we are working on it.” The board-ready answer takes a bit of preparation.

“What is our ROAI, and who owns it?”

Return on AI Investment is becoming a recognized metric in enterprise governance. The challenge is not the acronym. It is accountability. Most Series C companies have AI running across product, engineering, sales, and operations with no single person responsible for the aggregate return. That structure is common, but it means the answer to the board question is effectively “nobody tracks that in one place.” Naming an owner before the board does it formally is worth doing now.

“Are we in the 28% or the 72%?”

This specific framing will circulate. The board-ready answer does not require claiming to be in the top category. It requires demonstrating that a measurement architecture exists. The difference is between “here is what we track and why” versus “we believe it is working.” Boards at this stage are sophisticated enough to reward the former, even when the numbers are still maturing. I have seen founders earn board confidence not from strong ROI numbers but from the rigor of how they reported what they did not yet know.

“Are AI costs fully loaded in what we report?”

Most internal AI cost estimates are incomplete. They account for licensing and compute. They miss integration work, ongoing maintenance, change management, and the productivity cost of adoption curves. An analysis published April 8 found that enterprises underestimate their total AI costs by 40 to 60% on average. Boards at Series C do not expect perfect accounting. They do expect that someone is asking the full question.

The 60-Second Brief

If this needs to be addressed at the next board meeting:

“Our AI investments are producing [specific outcome: error rate reduction, cycle time, cost per transaction]. We track this against the baseline we set in [month]. Accountability for AI ROI sits with [name or role]. We have identified two areas where cost visibility is incomplete and are closing that gap this quarter.”

That is the whole brief. Boards at Series C are not expecting perfect AI accounting. They are expecting to see that AI deployment is governed, not just shipped.

The companies building AI measurement infrastructure now are not being cautious. They are building a governance advantage that shows up on term sheets.

What to Watch

The agentic AI layer adds urgency. Gartner’s research published this week included a forward-looking finding: 40% of current agentic AI deployments may be canceled by 2027 due to rising costs, unclear value, or poor risk controls. Series C companies scaling agentic workflows are building a future version of the same measurement problem, at higher cost and with more operational dependencies.

The governance discipline built around current AI programs is the foundation for whether agentic investments survive the next round of scrutiny. Making that connection explicit inside the company now, rather than after the board asks, is the practical move.