AI Employee Experience: 48% of Workers Fake Their AI Use

The data: almost half of all people are performing adoption

Research by Canadian software house Visier measures a threshold that every AI employee experience strategy has to confront: 48% of workers say they have overstated their AI use or their AI skills in front of colleagues and managers^[1]. The sample counts 1,000 full-time workers in the United States, working across healthcare, finance, human resources and other sectors.

The same group produces a second number, useful for reading the first: 54% say their role or career has changed substantially because of AI over the past two years. The original Visier research^[2] calls this behavior performative AI use.

The perimeter matters. These are people working in the United States, asked about self-reported behaviors and perceptions. The figure describes a specific population, and even on those terms it changes the value of every adoption dashboard hanging on an executive wall.

Adoption and competence remain two different measures

An adoption rate records how many people open a tool. Competence records how many people get a better result thanks to that tool. The two quantities are often read as one, and that is where the governance error begins.

When almost half the sample admits to inflating the story of their own use, the adoption rate loses its force as an indicator of real capability.

The point matters to anyone sitting on a board. An inflated metric produces investment decisions calibrated on competence that lives in slides rather than in workflows. The gap between reported use and actual use is where this challenge lives.

The distinction has to hold in internal language too. Saying that 80% of staff use a tool and saying that 80% of staff know how to use it well are two sentences with very different consequences for the budget.

The information vacuum that generates the performance

The same study measures the context in which the behavior arises. Nearly 60% of people say they have no idea what plan their organization is following to handle the changes AI brings.

A further 45% say they feel pushed to use AI even when they doubt their ability to use it effectively. Two figures describing the same condition: high pressure, thin information.

In that condition, displaying enthusiasm becomes the most rational individual strategy. The person reads a signal of expectation, perceives a risk to their own position, and produces the visible behavior that the signal rewards. The result is a designed organizational condition, more than an individual choice.

The mechanism: the incentive rewards the signal

The dynamic has a simple structure. Managers ask for AI use, measure reported use and reward whoever looks aligned. People respond to the measure, as happens with every indicator turned into a target.

The report traces the root of performative use to a deficit of trust in leadership. The formulation deserves attention: trust here functions as a structural variable, produced by the way managers communicate with, train and protect people during a transition.

Asking someone to "have AI do it" transfers onto them a translation task that belongs to the design of the work. The person receives a tool, an expectation and a deadline, while the method remains a private exercise. The performance fills that vacuum at zero cost to whoever created it.

The alternative reading, and how far it holds

One reasonable objection attributes the 48% to a problem of individual honesty. The reading has the merit of simplicity and the flaw of ignoring the two numbers that surround it.

A population informed about the company plan and trained on the tools would have little to gain from exaggeration. The behavior appears where information is missing and pressure rises, so it follows the condition more than the character.

There is a second objection, and a sturdier one: the research collects statements, so it measures perceptions. It is worth keeping in mind when the figure enters a board presentation. A self-report survey captures what people acknowledge doing, and 48% in all likelihood represents a lower bound, given how much it costs to admit to a performance.

What people are asking for, in their own numbers

The same research collects the demand for a remedy, and that demand turns out to be specific. 32% ask for more training and more AI development opportunities. 31% ask for greater transparency on AI-related matters.

The two requests correspond precisely to the two measured deficits: capability and information.

• Training: from the generic directive to hands-on practice on the real process
• Transparency: which roles change, on what timeline, with what effect
• Listening: concerns gathered before the rollout, never after

The detail has immediate operational value. People are pointing to the corrective lever with greater precision than many change management plans contain. The report urges managers to offer opportunities for learning and for exchange between colleagues, because that is how performative behaviors get prevented.

What the organizations closing the gap actually do

High-functioning organizations treat AI literacy as infrastructure, with learning paths placed inside working hours. They make the plan public: which processes change, on what timeline, with what effect on roles. They replace the usage metric with outcome measures, for example output quality and cycle time on a chosen process.

They also create spaces where admitting a difficulty carries no consequences, because declaring low familiarity is the first piece of data useful to designing the training.

One final practice separates those making progress: converting internal talent. Growing the people who already know the process produces applied competence faster than searching the external market, and it reduces the anxiety that feeds the performance.

The read for decision-makers

For the CEO, the conversation to bring to the board concerns the quality of the measure: which part of the reported adoption rate survives a check against outcomes. For the CHRO, the priority becomes a single line item: widespread literacy before any new release.

For the CFO, the investment with a documentable return is training hooked to the process, assessed on cycle time and on the quality of the work produced.

For the Talent & Compensation Committee, the human capital metric to track becomes the share of people who say they know their organization's AI plan. In Visier's US sample that value stays just above 40%. The design question for the next twelve months reads like this: what condition makes the truth about your own level of competence the most convenient answer for the person doing the work.