What both studies found
Microsoft published its Work Trend Index on May 5. Gartner released findings from its Global Labor Market Survey on May 13. The sample sizes are large: 20,000 workers across ten countries for Microsoft, 12,004 across forty for Gartner. The methodologies are different. And yet the core finding is nearly identical.
Microsoft puts it this way: organizational factors, meaning culture, manager support, and talent practices, account for 67% of reported AI impact. Individual effort accounts for 32%. The organization matters twice as much as the person.
Gartner arrives at the same place from a different direction: by 2027, half of enterprises without a people-centric AI strategy will lose their top AI talent. Not to AI itself. To competitors who figured out that the work is people, not platforms.
Both studies found that employees are already fluent. Microsoft reports that 86% of AI users treat output as a starting point, not a final answer. That is not a stat about hype. That is a stat about judgment. The people doing the work have developed a sense for what AI produces well and what it misses. Meanwhile, Gartner found that employees proficient with AI across multiple use cases are twice as likely to be highly productive and 3.2 times more likely to drive effective process improvements.
The fluency is real. The infrastructure to support it is not.
What caught me off guard was the anxiety data. Microsoft found that 65% of AI users fear falling behind professionally if they do not keep pace with AI. And yet 45% say it feels safer to focus on current goals than to redesign their work. Only 13% say they are rewarded for reinventing how they work with AI. Gartner’s data mirrors this: employees with a positive outlook toward AI are 3.4 times more productive, but that outlook depends on confidence in their current role and transparent communication about how AI will be used. Which most organizations have not provided.
(This is the part where, if you have led a team through any kind of technology change in the past five years, you are probably nodding. The people are not the problem.)
The gap between fluency and readiness
Both studies name a version of this gap, though they use different language for it. Microsoft calls it the “Transformation Paradox.” Gartner calls it the “Enablement Illusion.” Both are describing the same structural failure: leaders who mistake access for adoption, and adoption metrics for transformation.
Giving people a Copilot license is not the same as building the organizational capability to use AI well. And measuring how many people logged in last month is not the same as measuring whether AI is changing how decisions get made, how knowledge flows, or how work is structured.
We have a phrase we use internally at Ideal State: the difference between tools and capability.
Tools are what you buy. Capability is what sticks after the consultants leave.
Sara Teitelman
Most mid-sized companies and nonprofits we walk into have the tools. They have M365. They often have Copilot licenses sitting half-used. Some of their staff are quietly using ChatGPT and Claude on their own, without a policy, without guidance, without anyone asking what they are learning. Gartner found that 88% of employees with enterprise AI access also use personal AI tools for business tasks. That is not a rounding error. That is a workforce that has already started without organizational permission.
What they do not have is the infrastructure to make any of it compound. By infrastructure, I mean three specific things: governance that enables rather than restricts, knowledge management practices that give AI something accurate to work with, and managers who actively model AI use rather than delegating it downward.
Most mid-sized companies and nonprofits we walk into are at zero, one, or two of three.
Microsoft’s data confirms this from a different angle. When managers actively model AI use, employees report a 17-point lift in perceived AI value, a 22-point lift in critical thinking about AI output, and a 30-point lift in trust in agentic AI. Those are not small effects. But only one in four AI users say their leadership is clearly aligned on AI strategy. The managers cannot model what the organization has not decided.
Three moves in the next ninety days
The promise of this piece was practical next steps. Here is what we would recommend to any leader at a mid-sized company or nonprofit who reads both studies and wants to close the gap rather than study it further.
First, run an honest inventory of what is already happening. Not what your IT team reports is happening. What is actually happening. How many of your employees are using ChatGPT and Claude on personal devices, with company data, right now? The Gartner 88% number suggests the answer is most of them. The inventory is not a compliance exercise. It is the starting line for everything that follows.
Second, stand up lightweight governance in the first thirty days. Governance gets a bad reputation because most organizations implement it as restriction. What governance should do, when it is built well, is create freedom within a framework. A simple, published set of guidelines covering what AI can be used for, what data can go into it, and who to ask when the answer is not clear. This is a two-page document, not a six-month steering committee. It should exist within a month and be revised quarterly.
Third, give your managers permission and scaffolding to model AI use publicly. Microsoft’s data on the manager effect is the most actionable finding in either study. A 30-point lift in trust from manager modeling alone. That is not a training program. That is a cultural shift that starts with the leadership team using AI visibly, sharing what works and what does not, and making it safe for their teams to do the same. One standing thirty-minute session per week where managers share a single AI use they tried. That is enough to start.
None of these three moves requires a large budget. None requires a new technology purchase. All three require a decision from leadership that building organizational AI capability is the work right now, not something to revisit next quarter. The organizations that treat this moment as a people problem worth solving, rather than a technology project to manage, will be the ones whose advantage compounds over the next year. Both studies are saying it. The question is whether your organization will hear it as a research finding or as a starting line.