Why do employees keep turning to shadow AI?

Managing AI through IT alone is not a recipe for success — HR needs a seat at the table, expert says

Why do employees keep turning to shadow AI?

Since artificial intelligence (AI) deployment made its way into organizations, shadow AI and governance has been a constant conversation.  

And new survey data is showing the challenge continues, with many employees still using unsanctioned tools — pointing to a gap organizations need to close. 

The most common reason is speed. Forty-five per cent of workers say they use non-approved AI tools at least monthly to get work done quickly — and 53 per cent say it is because approved tools create extra work and friction, according to Ricoh’s 2026 Workplace Technology Adoption Gap survey. 

The survey also finds that 81 per cent of decision-makers admit their organization invested in new technology faster than employees could adapt. 

Emmanuelle Vaast, dean of the Desautels Faculty of Management at McGill University in Montreal, says the issue is primarily a gap between technology implementation and workforce readiness. 

“The emerging best practices show that AI should be managed not just by IT — that’s usually not a recipe for success — but by IT plus HR plus the CEOs, the executives,” she says. “If you have these three components, you have a better chance of a good implementation and use.” 

Why employees turn to shadow AI 

With many AI tools becoming more user-friendly and widely used in daily life, Vaast says it is not surprising that employees would also reach for them at work. 

Tools such as ChatGPT, Claude, and Gemini are ones many employees are already comfortable with — but they carry open risks to data and privacy leaks, unlike employer-approved enterprise tools whose security parameters are more controlled, she says. 

“From employees’ perspective, they’re thinking, ‘: I have a company-approved AI tool that is great and carries little risk to data leaks and privacy. But on the other hand, the more widely available tool might be easier and faster to use.’” 

Gap between IT and employee needs 

The gap between technology implementation and employee needs exists  because the approval of new tools is normally left to IT — who cannot always anticipate what employees need from technology, says Vaast. 

When it comes to approving new technology, IT tends to focus on security. That is why the decision needs to include HR professionals who understand the day-to-day work of employees, she says. 

“Involve HR professionals who know more about the work of employees and employees themselves to try and identify the needs — that would help close this gap.” 

That can be done through structured assessments. 

“Have a good communication from the start with employees to know what their needs are,” Vaast says. “Assess the needs, assess what the practices are, and from this, try to go with an articulation of what can be offered. Just because employees express a particular need for an AI tool doesn’t mean everything should be offered.” 

Balancing employee needs and security 

Even then, there is a balancing act, Vaast says.  

Approval of tools needs to be looked at through the lens of employee needs — but also security, and the impact on the broader chain of work those tools can create. 

“We’ve seen this happen over and over again — when they use an AI tool, they may do their work faster and more efficiently, but this might create more work down the line for other people: work checking, work transforming,” Vaast says. 

From an HR perspective, she says, the focus should not just be on the tasks individual employees do with AI — but on how those tasks affect the entire chain of work. 
A recent Anthropic report on AI and employment finds that while AI speeds up output, human attention quickly becomes the bottleneck — with employees spending more time monitoring and managing AI agents than on the work itself. The report finds 80 per cent of interactive clause session was spent idle waiting for human input.  

An AI tool might work better for some employees by getting their work done faster — but that does not mean automatic approval, Vaast says. 

“The assessment of needs and the assessment of practices are two huge steps to be able to design actual useful implementation for employees and for the work of the organization.” 

Latest stories