Businesses worldwide have invested billions of dollars into generative AI, yet recent headlines have raised questions about whether organizations are truly seeing returns. Despite this skepticism, the trend is not a retreat—it’s a shift toward refinement. Fewer companies expect to be “just starting” or stuck in pilots in 2026, while more are integrating AI into core operations, according to a recent IEEE global survey of technology leaders.

The report, “The Impact of Technology in 2026 and Beyond: an IEEE Global Study,” highlights increasing adoption of generative AI. Respondents were asked to identify the stage of AI adoption in their organization. Compared to 2025:

• Just Starting: 1% (down 4 points)

• High Expectations / Trying Small Projects: 8% (down 25 points)

• Challenged / Rethinking Approach: 4% (down 14 points)

• Learning / Seeing Some Benefits: 13% (down 11 points)

• Using Regularly, but Selectively: 39% (up 19 points)

• Rapidly Integrating / Expecting Bottom-Line Results: 35% (new in 2026)

A Period of Skepticism

New technologies often follow the “hype cycle”, a pattern where innovations surge in popularity, experience disillusionment, and eventually mature into practical value. Generative AI has been at the peak of expectations, but nearly one-third of projects are expected to be abandoned. McKinsey reports that while nearly eight in ten companies use generative AI, a similar proportion sees no measurable bottom-line impact.

“We’re entering a period of healthy skepticism following the natural technology adoption cycle,” said IEEE Senior Member Santhosh Sivasubraman. “This reflects a maturing understanding of AI’s capabilities and limitations. Companies that rushed deployments without planning are now reassessing approaches for more thoughtful implementation.”

Why Many AI Projects Fail

Survey data indicates a key challenge: overestimating AI reliability. The confidence of AI systems, particularly chatbots, often leads to unrealistic expectations. Half of the respondents cited over-reliance on AI and potential inaccuracies as top concerns.

“Inflated expectations have led leaders to imagine digital workers instead of systems designed for clear, measurable outcomes,” said IEEE Senior Member Eleanor Watson.

Common pitfalls include applying advanced models where simpler analytics suffice and using poor-quality data. Another frequent misalignment is confusing AI’s emotional appeal with functional value. According to IEEE Member Ning Hu, organizations often invest in chatbots or virtual assistants that sound friendly and empathetic but deliver little measurable impact.

Generating Real Value

Generative AI’s most reliable returns are emerging in operational back-end processes, Hu notes. Examples include:

• Automated order-to-cash workflows

• Inventory replenishment driven by predictive analytics

• Robotic process automation (RPA) for finance and HR

• Precision control in manufacturing or medical imaging

The survey identified cybersecurity, supply chain and warehouse automation, and software development as the top use cases for AI in 2026. Another resilient approach is sandboxed experimentation, where AI models explore creative solutions in isolated environments before outputs are transferred to production.

“In all these scenarios, AI’s value is measured in concrete, business-critical metrics, not user sentiment,” Hu said. “Sometimes the most valuable AI is surprisingly mundane—but highly effective.”

In short, the future of generative AI lies not in flashy applications, but in its ability to drive tangible, measurable business outcomes.