A new KPMG study has revealed that just seven percent of senior leaders report having established a return on investment from artificial intelligence, raising serious questions about how businesses are approaching AI adoption.
To discuss the findings, UC Today’s Christopher Carey spoke to Adam Hadley, Founder of AI consultancy QuantSpark, to explore why so many organisations are struggling to turn AI ambition into measurable results.
Magical Thinking Is Holding Businesses Back
According to Hadley, the root cause of failure often has little to do with AI itself. “A lot of that confidence is based on what I call magical thinking,” he explained, “that AI will just solve the intractable problems or challenges that a business might have.”
He pointed to Gartner’s hype cycle as a useful framework, noting that many businesses swing from extreme optimism to outright negativity when AI projects fail to deliver, often because they were not framed correctly from the outset.
Compounding the challenge is the rapid pace of AI development. Solutions that cost significant investment to build just a year ago can now potentially be replicated in a fraction of the time using the latest models, making it difficult for organisations to keep up.
The Importance of Systems Thinking
Hadley argued that successful AI deployment begins with a clear understanding of the business itself. Organisations need to map their core processes, identify where decisions are being made, and pinpoint the bottlenecks before introducing any AI solution.
“If you don’t know what functions your company has, the work they do, the value they’re providing, the bottlenecks, well AI isn’t going to solve that.”
Rather than large-scale transformation programmes, Hadley recommended a more targeted approach. Starting with one well-defined, high-impact process, demonstrating results, and then building from there allows businesses to develop trust and momentum gradually.
A Note on Cost Awareness
Hadley also flagged a growing concern around AI spend, warning that many teams are using expensive cloud-based models for low-value tasks. He pointed to open-weight models as a more cost-effective alternative that is gaining traction, though still maturing in terms of enterprise adoption.
