Sophisticated AI collaboration: An inside look at high-impact use

Analysis of 1.4 million AI interactions identifies the employee behaviors behind effective AI use—and how they can be taught at scale

Join KPMG and the University of Texas at Austin

Tuesday, March 24th, at 12 p.m. ET, for a LinkedIn Live discussion on our recent joint study.

The gap between routine and sophisticated AI use

As organizations accelerate artificial intelligence (AI) adoption, leaders are increasingly focused on a critical next question: is AI measurably improving the quality, speed, and ambition of people’s work? Who is succeeding with it, and why? What does successful use even look like? Many organizations struggle to distinguish routine AI use from the behaviors that translate into stronger judgment, more ambitious problem‑solving, and better work outcomes.

A joint study by KPMG LLP and the McCombs School of Business at The University of Texas at Austin helps clarify the gap. The researchers spent eight months studying KPMG back-office operations, analyzing more than 1.4 million real workplace AI interactions to understand how people use AI at work.

The findings reveal teachable differences between routine and sophisticated AI use that offer organizations a concrete road map for identifying and scaling high-impact AI capability.

To move beyond assumptions about what “good” AI use looks like, KPMG collaborated with Zach Kowaleski, Nick Hallman, and Jaime Schmidt, faculty members in McCombs’ Shulkin Department of Accounting, to analyze behavioral signals embedded in real-world AI interactions from over 2,500 employees—evaluating more than 30 characteristics of prompt behavior across months of usage data, including task complexity, prompting techniques, and iteration patterns.

What separated the best users wasn’t experience or technical know-how. This research surfaced consistent differences in how a small group of sophisticated users engaged with AI over time—how they framed problems, guided AI reasoning, and applied AI across complex tasks.

Sophisticated users treated AI as a reasoning partner, bringing it their most complex work and using it as a general cognitive tool rather than a narrow productivity aid. They set boundaries, specified structure, articulated objectives, and delegated cognitively demanding tasks across brainstorming, analysis, technical guidance, and problem-solving.

They guided the model’s thinking by asking it to assume a certain role or perspective, providing concrete direction and examples, showing it how to reason through tasks, and requiring it to explain responses. They pushed back on early answers and iterated for higher-quality results.

Regardless of role or level, effective users converged around four distinct behavioral patterns:

Frequency

Using AI as a routine part of daily workflow, not just occasionally

Persistence

Going beyond first answers, refining over multiple exchanges

Ambition

Starting with clearer, more detailed, and more complex prompts

Intentionality

Choosing the right tool or model based on task needs, not habit

The gap between routine and sophisticated AI use is not hidden in prompts themselves, but in patterns of engagement. And once those patterns are visible, they become possible to recognize, discuss, and scale. Iteration enables ambition, ambition drives strategic tool choice, and repeated success reinforces engagement.

Anu Puvvada

KPMG Studio Leader

Move from adoption to capability

AI adoption alone doesn’t guarantee effective AI use. What makes the difference is clarity and reinforcement: leaders setting a standard for sophisticated AI use and integrating it into how work gets done—just as KPMG has done for its own workforce.

At KPMG, this shift was driven by a firmwide training and enablement effort to help employees build the more sophisticated skills and behaviors identified in the research. Approximately 5% of users consistently demonstrated these behaviors across months of usage data, providing a clear, data-backed signal of what effective AI use looks like in practice. By embedding these behaviors into the firmwide learning ecosystem, more of the KPMG workforce can move from routine prompting to higher-impact human‑AI collaboration.

The same shift is achievable for other organizations when leaders deliberately create the conditions for these effective behaviors to take hold at scale:

Define what effective use looks like: Disseminate a clear definition of effective AI use—so expectations go beyond “use the tool.” Guidance can be tailored by roles and functions so employees understand where and how to apply AI in their work.

Train employees how to best work with AI: Provide resources such as practical playbooks, short explainers, and hands-on learning that focuses on how people actually work with AI. Give your people the space to practice, helping them move from familiarity to capability.

Reinforce high-impact behaviors: Build peer-led champion networks and communities of practice to amplify the behaviors that drive high‑quality outcomes. By showing what sophisticated AI use looks like in day‑to‑day work, these effective practices can spread organically across teams.

We realized early on that access to AI alone doesn’t drive better outcomes, a challenge many organizations are still grappling with. That’s why we put a deliberate set of AI‑enabled tools, training programs, and routines in place to make effective behaviors visible and expected, and to teach better problem framing, stronger supervision of AI, and purposeful iteration.

Steve Chase

US Vice Chair and Global Head of AI and Digital Innovation

✓ Build the habit: Encourage consistent use to build fluency and confidence. Show how AI can be integrated into daily workflows—from foundational tasks like drafting and summarizing, to more advanced collaboration such as pressure testing ideas and validating outputs.
✓ Push beyond first drafts: Normalize iteration with suggested follow-ups like “push this further,” “what’s missing,” or “rework for a different audience.”
✓ Raise the bar on prompts: Give examples of leading with context—audience, objective, and success criteria—and requesting structured outputs.
✓ Pause to match the task to the tool: Ask teams to consider, “What am I trying to accomplish—and what tool and approach best fits?”
When sophisticated AI collaboration is part of everyday work, organizations can move from uneven results to sustained performance improvement. This research shows that effective AI use is not rare intuition or individual talent; it’s a set of behaviors that can be cultivated at scale.

Access the study published in Harvard Business Review.

Ready to make the shift? KPMG can help guide you through the transformation required to accommodate hybrid AI and human workforces. Learn how you can empower your people to achieve more.

Learn more