Every day you open your laptop and hand ChatGPT or Claude an email draft, a line for your business plan, or some ad copy — and every so often, a question creeps in. Am I actually using this tool well, or am I just treating it like a fancier search bar? Office workers can at least glance over at a colleague's screen to see what prompts they're using, but solo founders running a business on their own have no one to compare notes with and no benchmark to check themselves against. So even though you're using AI every single day, there's a nagging sense that something's missing.
A recently published paper offers a clue. The study dug into a massive trove of actual prompts entered by back-office employees at a large corporation, using the data to show what separates skilled AI users from the rest within an organization. For solo entrepreneurs with no one to benchmark against, it's a rare glimpse into how other people actually use AI in practice.
The Study
Nicholas J. Hallman and three co-authors partnered with a large corporation to obtain eight months of real work data from 2025. They examined the full record — 713,564 prompts entered into generative AI tools, along with the responses, by roughly 4,000 back-office employees spanning 15 job functions. Because the researchers analyzed actual usage logs rather than surveys or self-reports, the findings capture how people really use AI, not how they say they use it. Their goal was to sift through this enormous pile of prompts to distinguish sophisticated use from unsophisticated use, and to pin down where that difference comes from.
What They Found
Analyzing the prompt data, the researchers identified three patterns.
What separated sophisticated use from the rest came down to three axes: seniority, job function, and time.
Starting with seniority, more senior employees showed a clear tendency to use generative AI in more sophisticated ways. The researchers attribute this to a complementary relationship between domain expertise and AI proficiency — in other words, using AI well requires knowing the job itself well, before any prompting skill even enters the picture. Turning to job function, the variation was substantial: departments that deal with company-wide direction and organizational change, such as strategy, digital innovation, and project management, showed the highest levels of sophisticated use. Finally, looking at the passage of time, employees' proficiency didn't improve on its own simply because more time went by, and even after formal AI training, the effect didn't last long. In other words, sophisticated use is not something that grows if left alone, nor something a single training session locks in for good.
Applying This to Your Own Work
Translated into terms useful for solo founders and planners, a few concrete moves emerge from this. If workplace AI proficiency ultimately comes down to job expertise, then knowing your own work deeply matters more than mastering prompt-writing tricks. Before polishing your prompting technique, it's worth checking how precisely you can define the problem you're actually working on. The finding that sophisticated use clustered around strategy and organizational-change work also suggests that AI makes a bigger difference when it's applied to direction-setting work rather than repetitive tasks — a good reason to lean on it more when refining a business plan or weighing a new direction. And the finding that proficiency doesn't stick after a one-off training session or tutorial applies just as much to solo founders. Rather than learning good prompting once and calling it done, it's better to keep AI woven into your actual work on an ongoing basis to keep that sense sharp.
A Few Caveats
This study looked at back-office staff at a single large corporation, so its findings can't be assumed to transfer directly to solo entrepreneurs working in an entirely different industry and organizational culture. What was confirmed is only the relationship between sophisticated use and seniority or job function — this study alone can't establish whether sophisticated use actually translates into better performance or revenue. Even so, the sheer scale of the dataset — more than 710,000 real prompts — and the finding that the gap between senior and junior employees isn't easily closed by training alone, give solo AI users good reason to focus their effort first on deepening their grasp of their own work.
