The Enterprise AI Wake-Up Call: What’s Driving Executive Skepticism?

The future of AI in the enterprise has become one of the most closely watched topics in business transformation. On the surface, investment is accelerating, and expectations remain high, but beneath that optimism, there are growing questions about whether organizations are actually seeing meaningful returns.

To explore this tension, Kristian McCann speaks with Laura Maffucci, Head of HR at G-P, who has been closely involved in new research examining how executives really feel about AI adoption across their organizations. Maffucci brings both data and practical HR insight into how AI is being deployed and where it is falling short of expectations.

Her perspective is grounded in recent findings that suggest a widening gap between investment and impact. The research sets the stage for a deeper discussion about whether organizations are deploying AI strategically or simply adopting it because of competitive pressure.

The ROI Gap And The “Supervision Tax”

At the heart of the research is a striking finding: many organizations are not seeing the returns they expected from AI investment. Direct in her assessment, Maffucci states that:

“AI really isn’t doing a good enough job yet,”

This shows a clear disconnect between ambition and outcome.

One of the most significant issues raised is what she describes as a “supervision tax” — the idea that AI outputs often require extensive human correction. She compares the experience to working with “a very junior level intern who’s producing something and giving it to you, and then you have to rework a lot of it.”

This inefficiency, she explains, is contributing to frustration at leadership level. According to the research, 73% of executives whose organizations have heavily invested in AI say at least some of that spending has failed to deliver ROI or meet initial goals.

Maffucci also highlights a deeper structural issue: many organizations are “chasing AI for the sake of AI.” Rather than identifying specific problems to solve, companies are layering AI onto existing processes without redesigning workflows, leading to added complexity rather than simplification.

Rethinking Deployment — From Tools To Transformation

Despite these challenges, Maffucci believes the issue is not AI itself, but how it is being implemented. She identifies three broad organisational behaviours: those aggressively “AI everything” without structure, those hesitant and unsure where to begin, and a third group attempting to implement AI in a more deliberate and system-wide way.

It is this third approach she sees as most effective. In her view, success comes from rethinking processes entirely rather than automating existing inefficiencies. “If we were to build this process now with the tools we have available, how would we build that?” she explains, emphasising the importance of redesign over retrofitting.

However, she acknowledges a major barrier: capability. Even when employees are willing to use AI, many lack the time or technical skill to build effective workflows. This creates a dependency on specialists or external support to implement AI properly, rather than expecting teams to self-serve.

Maffucci argues that organizations must therefore invest not only in tools, but in structure and enablement. Without this, she warns, AI risks becoming fragmented across departments, with inconsistent usage and uneven results.

Trust, Governance, And The Road To Confidence

The question of trust and confidence runs throughout the discussion. Maffucci argues that skepticism is not preventing adoption, but it is influencing how AI is used. Some organizations continue investing heavily, while others proceed cautiously or adopt tools without proper integration.

For Maffucci, restoring confidence depends on better governance of AI systems and stronger validation of data sources. She points to enterprise-grade tools, including GP’s own AI product GIA, which is designed around compliance and verified information. “You want to make sure that the information you’re getting is trusted,” she explains.

She also stresses the importance of understanding how AI systems are trained and what happens to company data once it is input. Without this clarity, organizations risk undermining trust and limiting long-term adoption.

Ultimately, the conversation points to a maturing phase of AI adoption. As Maffucci puts it, “you can hate AI or you can love AI, but it’s not going anywhere.” The challenge now is not whether to adopt it, but how to ensure it delivers value in a controlled, intentional, and sustainable way.