Shadow AI Is Testing IT Leadership. The Answer Isn't to Shut AI Down

TrustedTech is a leading Microsoft Cloud Solutions Provider (CSP) specializing in Microsoft Cloud support services, perpetual licensing, and comprehensive professional services for medium- and enterprise-sized organizations. TrustedTech is a member of the Microsoft Partner Network to ensure the delivery of industry-leading customer experiences. TrustedTech excels at helping IT teams navigate complex Microsoft 365 and Azure environments. Its teams across the U.S. and U.K are highly certified Microsoft architects and engineers who are among the world’s top experts, resolving advanced technical challenges faster than typical industry SLAs.

87% of IT and telecom workers say they’re confident using AI effectively, yet 68% are concerned about shadow AI. It’s a revealing paradox for an industry that should, theoretically, have a better handle on AI than most.

But confidence alone isn’t enough to mitigate the risks that come with it – AI confidence is an individual capability, AI readiness is an organizational one. The more confident employees become in testing the bounds of AI, the easier it is for adoption to move faster than the policies and organizational oversight designed to keep it in check.

No one in IT and telecoms is immune to shadow AI either. According to TrustedTech’s Shadow AI in the Workplace: The 2026 Research Report, senior decision-makers are more than twice as likely as their teams to use unapproved AI tools. On the surface, shadow AI is often treated as an employee problem, but the risk doesn’t disappear simply because someone has more experience under their belt.

When unapproved AI use stretches from employees to senior leadership, visibility has already started to break down. It’s not a secret AI is being used, but the real danger comes from not knowing which tools people are using, what they’re using them for, and what company or customer data is going along for the ride.

Bring Shadow AI into View

Before introducing new policies, IT and telecom leaders need to understand why people are working around the tools and policies already in place. Employees aren’t necessarily turning to unapproved AI because they want to break the rules – they may have found a faster way to solve a problem than the organization has given them.

Start with what’s already happening. Have open conversations with teams to find out where AI has found its way into their workflows and see what’s driving the gravitation towards those tools. Compare that information with the tools and access the organization already knows about, and you start to have a picture of what’s been happening out of view.

Pay attention if the same workarounds start showing up across teams. There may be a legitimate need that hasn’t been addressed yet. With that comes the opportunity to decide whether the use case can be supported safely, what guardrails it needs, and what should remain off limits.

Give AI Boundaries, Not Barriers

Bringing AI into an approved environment doesn’t mean giving it free rein. My recommendation is to start with the least amount of access a tool needs to be useful, then build from there. The closer AI gets to sensitive data or the systems that keep the business running, the stronger those guardrails should become. That could translate to tighter permissions, audit logging, and clear boundaries around what data AI tools can access.

But those guardrails still need to leave room for AI adoption and innovation. The two don’t have to sit at odds with security. There’s room to embrace both without sacrificing the human judgement that should sit alongside them. AI can take some of the heavy lifting off employees, but it still needs their expertise to give that work context.

Take a network issue. AI might help sift through logs and point a team towards a possible cause, but it won’t necessarily know what changed last week or what you might disrupt in the process of trying to fix it.

That same balance should extend to how employees experiment with AI in the first place. Give employees somewhere to go when they find a new tool or application worth testing, so that experimentation has a path forward without having to happen outside of the approved environment.

The aim isn’t to slow adoption in an industry where technology is constantly evolving and employees are confident enough to evolve alongside it. That confidence should be treated as an advantage, not a hindrance.

Train Beyond the Tools

Employee confidence doesn’t mean education and training completely disappear. And training doesn’t stop at showing an employee how to navigate a new tool and calling it a day. Training also needs to account for employees who are hesitant to use AI and those who become too reliant on it.

Human judgment should remain at the core. Teach employees to question what AI gives them, recognize where its limitations lie, and know when a task is better solved by their own expertise.

A one-and-done training is nowhere near enough either. As new tools emerge, so will the questions employees have and the ways they use them. AI adoption shouldn’t come at the expense of people developing their own skills. Employees still need to continue to grow their expertise alongside these tools, rather than gradually handing over more of their thinking to them.

Turn Adoption into Readiness

Shadow AI won’t be resolved by organizations writing stricter policies around it. IT and telecom employees are already confident with AI, and that confidence is only going to open the door to more ways of using it. Organizations don’t need to shut the door, but they do need to know what’s happening on the other side of it.

For more information, visit: https://www.trustedtechteam.com/

Netwrix Report Finds Midsized Businesses Face Higher Unauthorized Access Risks