Shadow AI: What it is and what HR & L&D can do about it

AI adoption has sped up across every sector. But for most companies, the technology turns out to be the easy part. The harder part is understanding how people actually use it.

Every organization has a list of approved AI tools. But most employees have a second, unofficial one they reach for whenever the sanctioned option feels too slow, too rigid, or too unclear to bother with.

That gap is what's known as Shadow AI, and it's growing faster than most companies realize.

Based on the most recent workplace studies:

• 52% of knowledge workers use artificial intelligence technologies that are not authorized by their employer.
• 65% of managers think their policy on AI is clear, while 57% of their employees consider it unclear or non-existent.
• 39% of workers using unauthorized AI technologies have entered confidential data from their company

‍

Why Shadow AI continues to flourish

In most cases, Shadow AI is not initiated as a result of circumventing policy.

All employees are trying to do is perform their work more effectively.

There are three common reasons for this trend.

‍

1. Approved tools do not always solve the actual issues at work

AI platforms are selected based on overall organizational requirements. But employees tend to face highly specific challenges.

In cases where the approved tools do not assist in doing those particular things, the need to look for other solutions arises automatically.

‍

“Teams can have access to AI without knowing how to use it in their day-to-day work ”

‍

notes Milica Sapic,, Talent and Organisational Development Manager at Publicis Sapient. This can leave employees more confused rather than more productive.

Technology adoption only succeeds when employees understand how AI supports their actual work, not simply because access has been granted.

‍

2. Organisations invest in technology more than people

Organizations channel most of their AI budget on software deployment. Much less is done to make sure that employees are able to use the software with confidence.

According to studies, organizations allocate about 93% of their AI budget to technology and only 7% on workforce enablement.

12% of the employees feel that they have been trained enough in AI, while 78% find training irrelevant to their job roles.

For HR and L&D, this reinforces an important insight: practical application drives adoption far more effectively than theoretical knowledge.

‍

“The most common mistake is starting with the technology,”

‍

notes Christopher Weggler,, People Operations & Innovation Director at commercetools. Employees don’t need to understand how AI works; they need to know how to use it effectively.

‍

3. Employees don't always feel comfortable asking

The biggest problem does not always lie with technology itself. Rather, it is about psychological safety.

When employees are not sure if the AI application can be used appropriately, they tend not to ask for information due to fear of appearing uninformed.

‍

“To learn something new, people need to feel safe asking questions and making mistakes,”

‍

notes Daria Rudnik, Team Architect and Partner at AI Leader’s Compass.

Creating psychological safety is therefore just as important as creating AI policies.

‍

‍Shadow AI happens when employees work around existing policies, use unapproved tools, and receive limited guidance on how to use AI effectively. Training is often one-off, while AI use remains hidden rather than openly discussed.

Healthy AI adoption looks different. Employees understand the organisation’s AI policy and have access to approved tools that match their needs. Managers actively coach AI use, learning continues over time, and employees feel comfortable discussing how they use AI and sharing feedback.

‍

What HR & L&D can do about Shadow AI

Rather than approaching Shadow AI as simply a compliance problem, businesses need to view it as a learning problem.

Three things have the most impact.

1. Integrate AI learning into day-to-day operations: Learning through exercises tied to employees actual tasks works far better than generic AI classes. People pick up new tools faster when they're using them to solve a problem they already have.‍
2. Make it easy to ask:Make sure that there are ways for employees to find out whether an app would be right to use.That could include a simple AI help desk, a dedicated Slack or Teams channel, or a clear process for requesting approval for new tools.If asking is easier than working around the policy, employees are far more likely to follow it.
3. Prioritize managers:Managers actions shape how their teams use AI day-to-day. They decide which shortcuts are acceptable and which tools are worth adopting. That makes manager training one of the most effective levers HR and L&D have for responsible AI adoption.