Aggregate adoption tells you who used AI. It doesn’t show who has integrated it deeply enough to save time, expand what they can do, and build repeatable workflows.
Key Takeaways
- OpenAI found frontier workers send 6x more messages than median users, with a 17x gap for coding. That shows how uneven AI use can be, but not whether more activity produces better work.
- OpenAI also found that workers using AI across roughly seven task types reported 5x more time saved than workers using it across about four. Breadth and depth of use are more informative than access alone.
- The goal is to identify employees whose AI use produces strong outcomes, understand which tools and workflows matter, and scale the practices that hold up.
What the 6x Gap Actually Measures
OpenAI’s enterprise data shows a major difference in usage depth. Frontier workers, defined as the 95th percentile of adoption intensity, send 6x more messages than the median worker, and the gap reaches 17x for coding messages.
That doesn’t mean frontier workers are six times more productive. Message volume measures activity, not output quality, time saved, or business value. A heavy user may have built valuable workflows or may simply be generating more activity.
The stronger finding connects breadth of use with reported time savings. Workers who use AI across roughly seven task types reported 5x more time saved than those using it across about four. Employees who reported more than 10 hours saved per week also used multiple models, more tools, and a wider range of tasks.
Leaders need to move beyond asking who sends the most prompts. The useful question is whose usage patterns are producing measurable value.
Why Adoption Rates Hide Your Power Users
Aggregate adoption counts participation. It can’t show the difference between an employee who opens an AI tool once a week and one who has integrated it into their daily workflow.
Larridin customer data shows how wide that gap can be. In one environment, ChatGPT desktop users generated 130 events per user, compared with 18 events per browser user. That’s a roughly 7x difference in engagement depth hidden inside the same adoption count.
Even among active users, OpenAI found significant gaps in feature use. Of monthly active ChatGPT Enterprise users, 19% had never used data analysis, 14% had never used reasoning, and 12% had never used search. An active-user metric can include people with very different levels of feature adoption and workflow integration.
Different Studies Point to the Same Management Question
Other research finds stronger reported outcomes among advanced users, but the groups and measures aren’t interchangeable.
Microsoft’s 2026 Work Trend Index found that 80% of its Frontier Professionals reported producing work they couldn’t have produced a year earlier, compared with 58% of all surveyed AI users. Microsoft defines Frontier Professionals based on advanced agent use, workflow redesign, and repeatable AI-enabled practices.
Writer’s 2026 vendor survey used the term super-users. They reported nearly 4.5x more weekly time savings than lower-adoption AI users, while 87% of leaders said super-users were at least 5x more productive. Those are self-reported and leader-perceived outcomes, not direct productivity telemetry.
The studies use different definitions, but they lead to the same practical question: can you identify your most advanced AI users and verify what their behavior is producing?
What to Measure Beyond Usage Volume
Finding power users requires more than ranking employees by messages or sessions. Leaders need to examine four dimensions:
- Usage depth: How consistently does the employee use AI and across how many tasks?
- Capability breadth: Which models, features, tools, and workflows do they use?
- Workflow integration: Has AI become part of a repeatable process rather than an occasional shortcut?
- Outcomes: Does the work save time, improve quality, reduce rework, or create measurable business value?
Larridin’s AI Fluency platform segments users by adoption stage, identifies advanced users and power teams, and shows how AI is being used across functions. Pairing those behavioral patterns with AI proficiency and outcome data helps separate valuable power users from employees who are simply active.
High activity is a clue. Strong outcomes are the evidence.
Turn Power-User Workflows Into Internal Curriculum
Once leaders identify employees with high AI proficiency, the next step is to study what they do differently.
Look at the specific tasks, features, models, prompts, agents, and workflow changes behind the results. Then test whether those practices transfer to other employees doing similar work.
This creates a stronger training foundation than generic prompting advice. The examples come from your organization, use your systems and data, and have already produced observable results in a real operating environment.
Not every power-user habit should be scaled. A workflow may depend on unique expertise, create hidden review work, or perform poorly outside one person’s role. Validate quality, risk, and business impact before turning an individual practice into a company standard.
Frequently Asked Questions
Who are AI power users?
AI power users are employees who use AI frequently and deeply across tasks, features, and workflows. Usage alone isn’t enough. Leaders should also consider whether that activity produces time savings, better output, less rework, or other business value.
How do we identify AI power users in our organization?
Start with behavioral data such as usage consistency, task breadth, advanced feature adoption, and workflow integration. Then connect those patterns to outcomes. This prevents high activity from being mistaken for high performance.
Does more AI usage mean higher productivity?
Not automatically. OpenAI’s 6x and 17x findings measure message volume. Its separate time-savings data suggests deeper and broader use is associated with stronger reported benefits, but organizations still need their own outcome data.
What should we do after identifying power users?
Document the workflows that produce value, test whether they work for others, and build role-specific training, prompt libraries, templates, or mentoring around the practices that transfer. Continue measuring quality and outcomes after scaling them.
Find the Power Users Hiding Inside Your Adoption Rate
Larridin helps leaders identify advanced AI users and power teams, understand what they do differently, and connect those patterns with measurable outcomes.
Book a discovery call to see where advanced AI usage is producing value and where the rest of the organization needs support.