Intro:
The paper, "AI in the Enterprise: How People Use M365 Copilot Chat," provides one of the clearest views yet into how artificial intelligence is being integrated into the daily workflows of millions of employees across more than a million companies, including nearly 70% of the Fortune 500. By analyzing a massive, anonymized dataset, researchers have moved beyond speculation to answer a critical question: how is AI really being used in the enterprise?
What Are People Using AI For?:
The research team classified millions of user prompts into distinct "intents," revealing the most common ways people turn to their AI assistant. The analysis shows that two categories dominate, accounting for nearly 60% of all interactions: Information Inquiry (asking questions) and Content Refinement (editing and improving existing text)
These are followed by other crucial work activities like generating new content, programming assistance, and analytical reasoning. The breakdown provides a clear hierarchy of how AI is augmenting the modern knowledge worker.
A visual breakdown of M365 copilot usage
Perhaps the most significant finding is the "broad but uneven" pattern of adoption. Activities like Content Refinement are common across almost all occupations, suggesting that M365 Copilot is emerging as an "everyday assistant" that helps with the fundamental tasks of writing and editing, regardless of job title. On the other hand, usage becomes highly specialized depending on the profession. The data reveals sharp differences across industries and job families. For example:
Programming Assistance is, unsurprisingly, concentrated among technical roles.
Some occupations now use Copilot more for Content Refinement than for basic Information Inquiry, indicating a shift in how they perceive the tool's value
This dual nature shows AI acting as both a general-purpose utility and a powerful, role-specific specialist tool.
The Next Frontier for AI in the Enterprise
The study also points to a potential evolution in user behavior: a subtle shift away from "chat as search." Researchers observed a 5% relative drop in the share of "Information Inquiry" over the 114-day study period, suggesting that as users become more sophisticated, they move beyond simple questions toward more complex content and communication-related work
More importantly, the "uneven" adoption highlights the next frontier for AI. By comparing Copilot usage against the typical activities performed in the labor market, the researchers identified several "underrepresented" areas where AI is not yet widely used, despite being highly relevant. These include:
Documenting and Recording Information
Work activities related to evaluating the quality or accuracy of data, particularly in the Banking industry
A significant portion of tasks common in the Consulting industry
These gaps don't represent failures, but rather the largest opportunities for growth. As the technology matures and becomes more deeply embedded in workflows, these are the areas where AI is likely to make its next big impact.
The study concludes that enterprise AI is past the "novelty stage" and has become a substantive part of day-to-day knowledge work. The story of AI at work is no longer about what it can do in principle, but what it is doing in practice—and where it will go next
Reference:
AI in the Enterprise: How People Use M365 Copilot Chat
Top comments (0)