<p>A study of 1.4 million real workplace interactions with artificial intelligence reveals clear, reachable differences between routine and sophisticated Al use — offering organisations a practical roadmap for sealing high-impact Al capability.</p><p>Conducted by KPMG LLP and the McCombs School of Business at The University of Texas at Austin, the research identifies observable patterns in how top users frame problems, guide AI reasoning, and apply AI to complex tasks. The findings, published in Harvard Business Review, are already shaping KPMG's internal practices and client work.</p><p>Over eight months, researchers analysed Al use across KPMG's back-office operations. The most effective users labelled "sophisticated were not simply frequent users or technical experts. Instead, they stood out for how they engaged with AI: framing problems clearly directing the moder's approach, and integrating AI into brooder workflows.</p><p>To define "good" AI use, researchers examined more than 30 behavioural signals, including task complexity prompting techniques, and iteration patterns. Rather than identifying abstract "power users”, they focused on individuals who had learned to think with the model, not just query it.</p><p>Sophisticated users consistently treated Al as a reasoning partner: They assigned roles or perspectives, provided clear instructions and examples, and required explanation for outputs. They refined responses through multiple iterations and applied AI to complex, high-value tasks. They also set boundaries, defined structure, and used AI across brain-storming, analysis, and problem solving — treating it as a general cognitive tool rather than a simple productivity aid.</p><p>These behaviors produced measurable patterns. Sophisticated use correlated with four signals: frequent return to AI, persistent refinement of outputs, ambitious initial requests, and intentional tool selection. Only about 5 per cent of users consistently demonstrated these traits</p><p>KPMG has translated these insights into firmwide training and enablement programs. By embedding Al-first behaviours into learning systems — through structured trainings, playbooks, and peer networks — the firm is helping employees wow from basic prompting to high-impact collaboration.</p><p>These same principles now guides how KPMG supports clients in defining effective AI use, building role specific capabilities, and scaling human-Al collaboration across everyday work.</p>
<p>A study of 1.4 million real workplace interactions with artificial intelligence reveals clear, reachable differences between routine and sophisticated Al use — offering organisations a practical roadmap for sealing high-impact Al capability.</p><p>Conducted by KPMG LLP and the McCombs School of Business at The University of Texas at Austin, the research identifies observable patterns in how top users frame problems, guide AI reasoning, and apply AI to complex tasks. The findings, published in Harvard Business Review, are already shaping KPMG's internal practices and client work.</p><p>Over eight months, researchers analysed Al use across KPMG's back-office operations. The most effective users labelled "sophisticated were not simply frequent users or technical experts. Instead, they stood out for how they engaged with AI: framing problems clearly directing the moder's approach, and integrating AI into brooder workflows.</p><p>To define "good" AI use, researchers examined more than 30 behavioural signals, including task complexity prompting techniques, and iteration patterns. Rather than identifying abstract "power users”, they focused on individuals who had learned to think with the model, not just query it.</p><p>Sophisticated users consistently treated Al as a reasoning partner: They assigned roles or perspectives, provided clear instructions and examples, and required explanation for outputs. They refined responses through multiple iterations and applied AI to complex, high-value tasks. They also set boundaries, defined structure, and used AI across brain-storming, analysis, and problem solving — treating it as a general cognitive tool rather than a simple productivity aid.</p><p>These behaviors produced measurable patterns. Sophisticated use correlated with four signals: frequent return to AI, persistent refinement of outputs, ambitious initial requests, and intentional tool selection. Only about 5 per cent of users consistently demonstrated these traits</p><p>KPMG has translated these insights into firmwide training and enablement programs. By embedding Al-first behaviours into learning systems — through structured trainings, playbooks, and peer networks — the firm is helping employees wow from basic prompting to high-impact collaboration.</p><p>These same principles now guides how KPMG supports clients in defining effective AI use, building role specific capabilities, and scaling human-Al collaboration across everyday work.</p>