The AI Fluency Gap
Why Access to AI Is Not the Same as Value from AI
Three years after the widespread arrival of generative AI in the workplace, a consistent finding is emerging across multiple independent research programs: most organizations have access to AI tools, but very few have employees who use them in ways that generate measurable outcomes. The gap is not technological. It is human.
A Google/Ipsos study published in February 2026 found that only 5% of workers meet the definition of “AI fluent” – meaning they have meaningfully redesigned how they work using AI, not merely used it for occasional tasks.1 That 5% cohort is 4.5 times more likely to report higher wages and 4 times more likely to report a promotion directly attributed to their AI capability. The remaining 95% are, at best, using AI as a faster version of their existing workflows – search, summarization, email drafting – without the deeper integration that generates competitive advantage.
Independent research from the University of Texas at Austin and KPMG, analyzing 1.4 million real workplace AI interactions over eight months, reached the same conclusion from a different angle.2 Sophisticated AI use – the kind that actually moves performance – was not predicted by frequency of use or technical skill. It was predicted by specific patterns of engagement: how users frame problems, guide the model’s reasoning, iterate deliberately, and apply AI across complex cognitive tasks rather than narrow ones. Approximately 5% of employees demonstrated these patterns consistently.
A third body of research from the NeuroLeadership Institute adds a cognitive dimension to these behavioral findings.3 The characteristic that most reliably separates genuinely fluent AI users from everyone else is metacognition – the habit of thinking about one’s own thinking before, during, and after an AI interaction. Fluent users approach AI with three specific habits: humility, flexibility, and vigilance. These habits are not innate traits. They are trainable skills – and the research is explicit that organizations which develop them deliberately outperform those that simply provide tool access.
Taken together, the research makes a clear case: the bottleneck in AI value creation is not the technology – it is the thinking behind how people engage with it. Organizations that address this through structured development of AI fluency, engagement patterns, and metacognitive habits consistently close the gap between adoption and outcomes. Those that treat AI as a self-service productivity tool and leave employees to figure it out independently are, according to the data, leaving the majority of potential value unrealized.
This is precisely the gap that B-Sharp AI’s Training & Advisory and AI Getting Started & In-Housing service lines are designed to close – not by teaching people which buttons to press, but by building the habits, frameworks, and patterns of engagement that turn AI access into AI capability.
1 Google/Ipsos (2026). AI Fluency in the Workforce. Reported in Fortune, February 19, 2026.
2 Hallman, N., Kowaleski, Z., Puvvada, A., & Schmidt, J. (2026). What the Best AI Users Do Differently. Harvard Business Review, March 2026. Joint study with KPMG LLP and University of Texas at Austin.
3 Rock, D., & Weller, C. (2026). The One Skill That Separates People Who Get Smarter with AI from Everyone Else. Fortune, March 21, 2026. NeuroLeadership Institute.