<img height="1" width="1" style="display:none;" alt="" src="https://dc.ads.linkedin.com/collect/?pid=306561&amp;fmt=gif">
Skip to content

Is AI Meeting ROI Expectations?

Our recent research found that AI is already meeting or exceeding initial ROI expectations for many leaders. However, further results may depend on technical and organizational infrastructure that many are yet to complete. 

 

Early AI investments appear to be paying off for most organizations. Of the 1,004 leaders we surveyed, 85.1% say AI has met or exceeded their initial ROI expectations, with just 14.9% reporting that it fell short.

But delivering on expectations is only half of the battle. Meeting today’s expectations is one thing, building the infrastructure to scale that success for the future is another, and that’s where the gap is starting to show.

Our research suggests that, for many organizations, the technology needed to unlock the next phase of AI-driven ROI is not missing but unfinished. Scaling that value will require stronger infrastructure, closer cross-functional alignment, and more AI upskilling for their staff. What’s more, almost all leaders think closer collaboration between AI and technology leadership on a shared deployment platform could unlock further gains (96.4%). 43.3% expect gains of up to 10%, 41.9% expect 11% to 25%, and 11.2% think gains could exceed 25%.

Opportunity for Growth in AI Decisioning

While the industry has seen great strides in overall adoption and returns, the real opportunity resides in operationalizing AI decisions at scale. Today, AI drives only a small share of the decisions that actually reach customers, just 28.5%, split between traditional machine learning (13.1%), generative AI (8.9%), and agentic AI (6.5%). The remaining 71.5% are still made by humans, rules-based systems, or standard analytics. That gap, between strong returns where AI is used and how rarely it’s used yet, is where leaders expect the next wave of ROI to come from.

Turning AI Capability into Measurable Returns

Dr. Scott Zoldi, Chief Analytics Officer at FICO, notes that closing that infrastructure gap is one of the most consequential actions enterprises can take to improve ROI. Consistency, more than any individual model, is what compounds into ROI at scale.

“Organizations that follow a Responsible AI framework produce far more consistent models,” he says. “That consistency gets models to market faster, and that’s a huge ROI for an organization: not just time to market, but time in market.”

Closing the infrastructure gap is only part of the equation. Madhu Mukherjee, VP and Head of Transformation and Data Science at USA TODAY Co., a leading US news and media company, points to what she calls the ‘last-mile problem.’

“Organizations invest heavily in developing or acquiring an AI capability, but less attention may be given to the final steps: who will use the output, what decision it will change, how quickly someone can act on it, and what happens when the output is wrong,” she says. “That is often where the expected value disappears.”

She also names a related discipline worth building alongside execution, strategic alignment: “Organizations can be too quick to make this a technology question,” she says. “The model may be doing exactly what it was designed to do, but that does not mean it is solving a problem worth solving or solving it at a cost that makes commercial sense.”

Mukherjee notes that investing in more capable technology will create value only if it is directed at the right problems and embedded in decisions people can act on.

This distinction will become even more important as organizations decide where to focus their AI investment next.

Domain-specific language models (49.5%) and generative AI (48.7%) top the list of technologies expected to drive ROI over the next 18 months, narrowly ahead of interpretable model architectures (46.6%) and Responsible AI standards (42.9%).

Rachael Hadaway, Vice President of AI Product at FICO, says specific capabilities must be embedded at the architecture level to make a shared platform effective.

“Tech leaders need to think beyond a shared dashboard. A shared platform should operate as a unified decisioning engine,” she says. “When a model in fraud and a model in originations both run through it, they must inherit the same audit trail, the same monitoring, and the same explainability standard. A shared platform makes this process seamless and automatic. Without the right infrastructure in place, organizations will find themselves rebuilding governance for every new use case.”

How Leaders Measure What’s Working

Our research shows that when leaders measure ROI, they often lean on a broad, largely outcome-oriented set of metrics rather than pure financials.

Customer satisfaction scores are the single most common measurement, cited by 62.7%, ahead of revenue impact tracking (58.6%), operational efficiency benchmarks (54.4%), and cost reduction metrics (46.2%). Only 1.8% say they do not formally measure AI ROI at all.

Mukherjee cautions that these metrics can overstate AI’s contribution if they capture improvements in one part of a workflow without accounting for their wider effects.

“Another important distinction is whether the initiative improves the entire decision process or simply optimizes one step,” she says.

“AI can make one part of a workflow more efficient while shifting cost, risk, or complexity somewhere else. Material improvement should therefore be measured end to end, including the effort required to review outputs, handle exceptions, and correct mistakes.”

What’s Still Getting in the Way?

In addition to addressing the technical infrastructure gap that exists, enterprises must also create the organizational alignment needed to make AI implementations effective enterprise-wide.

Technical barriers aside, our research found that the biggest barrier to AI adoption is cultural. Asked what’s preventing alignment on AI initiatives, 61.4% of leaders cite resistance to change, ahead of a lack of shared understanding of AI, at 58.4%. Both outrank resource limitations (39.8%), weak business-IT collaboration (35.2%), and a lack of unified AI strategy (33.9%). Leadership buy-in is the least-cited challenge, at just 15.8%, suggesting leaders locate the problem in the broader organization rather than their own ranks.

Kajal Chokshi, who leads Strategic Workforce Planning, AI & Automation globally at Mondelēz International, a multinational snacking and food company, points to a specific, practical source for resistance to AI: “The biggest reason I see people feel this way is this fear that they are going to be replaced by AI,” she says. “That’s a very legitimate fear.”

Her prescription starts with education: building “centers of excellence” around what AI can and can’t do, before asking whether it should be used for a given task at all.

Haroon Abbu, SVP, Digital Technology, AI & Data Analytics at Bell and Howell, a major US business process automation company, ties this to leadership fundamentals rather than technology. “It comes down to people, process, and technology,” he says. “In isolation, technology is not a sufficient solution. It really comes down to change management and leadership transparency.”

_

This article is a preview from our latest research report published with FICO. Check out the full report here: The Confidence Gap in AI Decisions. 

Download the full report