Missing AI ROI Traces Back To Ungoverned Value Adoption And Decisions

Somewhere this quarter, an AI governance committee walked through model risk, data quality, and hallucination rates, and everyone left the room satisfied. Nobody in that room owned whether the AI created value, whether anyone adopted it, or whether a single decision improved. Meanwhile, 78 percent of executives doubted they could pass an AI governance audit, the returns stayed missing, and the two facts looked related.

Dr. Elena Alikhachkina is a four-time Fortune 500 Chief Data Officer, board director, and founder of InfoFluence™. She is the author of AI Oversight: A New Mandate for Corporate Directors and Executives, a book informed by interviews with more than 150 corporate directors and senior executives. A PhD economist and globally recognized AI, data, and digital transformation leader, Alikhachkina has served in executive roles at companies including Danone and Roche. Today, she advises boards and executive teams across healthcare, manufacturing, consumer products, and software on AI governance, data strategy, and digital transformation. Through her work with directors and executives, she repeatedly encountered the same challenge: despite significant investments in AI, many boards lacked a practical framework for effective oversight and governance.

"Governance must include the value, the governance of adoption, and the decisions. It should focus on the creation of value and adoption, because every single enterprise is struggling with return on investment," Alikhachkina said.

• An operating model is not governance: Aimed low and the org chart emptied of owners. "The first risk is nobody has accountability for anything. If you start by governing value creation, you appoint somebody accountable for decisions. If it's about a tool, who is accountable for delivering value from this tool?" Customer experience showed the cost, even as enterprises added real-time controls to protect customers. "Without governance, people measure what was removed. Call time decreased, no personal interactions anymore, big win, we saved three dollars. But nobody's tracking what I lose in customer equity and customer engagement," Alikhachkina said.

The membership of the small club seeing real AI returns did not grow on technology briefings. "The only way is to educate executives, not about AI, but how to do business with AI. All the education executives get right now is about the technology. You need to show people how to do business differently if you have the technology," noted Alikhachkina.

• Persuasion took numbers: "Going bottom up and saying we need governance never works," said Alikhachkina. "I was explaining to an executive officer how much business he was losing, and he said, nope, I don't believe you. Show me. So I partnered with the commercial officer, we manually crunched the numbers, and I showed him that just in Europe, last year, you lost 120 million because you didn't have governance and didn't know where the money was going." In her view, the skill was scarce. "The majority of people accountable for governance right now don't have the commercial skill. They cannot connect governance to a business case."

• The commercial model: Alikhachkina's proof point predated the 95 percent of pilots that never reach the P&L. "Roche has been one of the champions in healthcare in changing the commercial model. When you change the commercial model, you immediately change accountabilities and governance. They didn't do it by implementing technology."

• Closed the loop: The analytics engine her team built delivered tens of thousands of personalized recommendations to the field, and human validation closed the loop. "We validated recommendations with people. We asked, do you believe this recommendation? People rated them, and 80 to 86 percent were rated as relevant." Domain knowledge made the difference. "Somebody can recommend a promotion in Walmart, but there's knowledge sitting with people. Walmart runs promotions every four weeks, BJ's every three weeks. AI doesn't know this."

Vendors now write governance into their AI principles, and a growing camp treats it as a strategic advantage rather than brake pads. "AI governance is going to grow. It's one of the biggest growing areas, and my projection is it becomes a business discipline, not a technical discipline. Because of the missing return on investment and missing adoption, companies are waking up."

The pressure lands on boards, where two thirds of directors admit limited to no AI knowledge. "Many boards do not have digital expertise. Countries outside the US already have regulations that boards must have digital expertise," Alikhachkina concluded. "The US doesn't, so I think the US is waiting for the lawsuit to happen before we get regulation in place."