Your Employees Are Already Using AI. Do You Know How, Where, or Whether It’s Safe?

This article, which expands on insights from a recent Solutions Spotlight event with Nexthink, explores how companies can ensure secure employee use of AI, equip teams with the right tools at the right time, and track adoption rates.

The AI adoption story inside most enterprises right now has two versions. The version leadership sees: a curated rollout of approved tools, a training program, a licensing agreement, and a policy document. And then there’s the version that is actually happening, where employees use whatever AI tools work best for them, regardless of whether those tools appear on any approved list, while the tools the organization paid for sit significantly underutilized.

Closing that gap requires visibility, and most organizations lack it. That is the starting point for everything Nexthink Senior Product Marketing Manager Shawn Lazarus and Solution Consultant Kevin Satterwhite walked through during a recent Solutions Review Solution Spotlight, covering how enterprises can gain real-time insight into AI tool usage, enforce governance without creating friction, and turn adoption data into a practical, iterative AI strategy.

What AI Adoption Visibility Actually Looks Like Across an Enterprise

The numbers framing this conversation are striking on their own. According to Gartner, “Over 40 percent of agentic AI projects will be canceled by the end of 2027, due to escalating costs, unclear business value, or inadequate risk controls.” On top of that, 95 percent of enterprise AI pilot programs deliver no measurable financial return, according to research from MIT and Fortune. And within organizations that have deployed AI tools, a third of employees are unaware that the tools are even available to them.

That last figure is the one that Lazarus and Satterwhite return to most pointedly. Organizations cannot derive ROI from tools employees do not know exist, do not understand how to use, or have quietly decided to avoid because the surrounding policy is unclear. The adoption gap is not primarily a technology problem. It is a visibility and communication problem that compounds over time as the AI tool landscape expands faster than any training curriculum can keep up with.

How to Turn AI Usage Data Into a Companywide Strategy

The foundation of the Nexthink approach is what Satterwhite calls turning the lights on: giving leadership a real-time, consolidated view of how AI tools are actually being used across the organization before trying to optimize, govern, or expand that usage. The platform surfaces weekly AI usage per employee, tracks seven-week trends to show whether adoption is growing or plateauing, breaks usage down by department, and compares organizational patterns against peer benchmarks. That last capability matters because, without context, a number like ninety minutes of weekly AI usage per employee is meaningless. When benchmarked against the industry, it becomes actionable.

Department-level breakdowns are where strategy conversations get specific. Satterwhite described the analytical posture this creates for AI leaders: when one department shows significantly higher adoption than others, that is not just a data point. It is a playbook waiting to be extracted. Reaching out to the high-adopting team to understand what drove their success, then using that insight to lift teams that are struggling, turns usage data into an internal best-practice library that no external consultant can provide.

The same dashboard surfaces tool-level insights. If ChatGPT is leading in active users but Copilot shows higher usage intensity among the smaller group that uses it, that tells a different story about where each tool fits in the workflow. Some employees need a general-purpose tool. Others have more complex use cases that benefit from deeper integration. Visibility into that distinction lets strategy move beyond blanket rollout decisions.

How Governance and Compliance Work Without Killing Adoption

The compliance dimension of AI governance is where many organizations reach for blunt instruments: block unsanctioned tools, mandate approved ones, circulate a PDF policy document. The problem with that approach is not that it fails to communicate the policy, but that it communicates the policy at the wrong moment, in the wrong place, and without giving employees a viable path forward.

Instead, companies should aim to intercept employees at the point of behavior rather than in a training session weeks earlier. When a user navigates to an unsanctioned AI tool, they receive an in-browser notification explaining that the tool does not meet the organization’s policy and are redirected toward the approved alternative. The message arrives at the exact moment it is relevant and includes a clear next step rather than simply blocking access.

For approved tools, the same just-in-time delivery mechanism works in reverse. A user accessing Copilot for the first time can be greeted with a guided walkthrough that meets them at the point of intent, when they are actively trying to accomplish something, rather than expecting them to remember training they completed during onboarding. Organizations that want to support employees who are more comfortable with one tool can build guides that explicitly show how to accomplish familiar workflows in the company’s preferred platform.

Policy reminders can be configured to appear at contextually relevant moments, such as when an employee attempts to attach a file to a conversation, surfacing a reminder about what data classifications are acceptable in that specific tool. The frequency of these reminders is configurable, so organizations can balance awareness with the risk of training their employees to dismiss notifications reflexively.

Identifying the Adoption Gap Before It Becomes a Cost Problem

One of the most concrete financial applications of AI adoption visibility is license optimization. Most enterprises are simultaneously overspending on AI licenses held by employees who never use them and underspending on higher-tier access for employees whose workflows would benefit from more advanced capabilities.

Nexthink’s metering approach uses a 90-day window to identify inactive licenses. A user who has not engaged with a licensed AI tool in that period becomes a candidate for license reclamation. The reclaimed license can be reallocated to someone who has been requesting access. If the original user eventually wants access again, an automated workflow handles the request, routes it through the appropriate approval chain, and restores access without requiring the employee to navigate a manual IT process.

The same workflow logic applies on the other end of the licensing spectrum. When an employee encounters a feature that requires a higher license tier, an in-app notification can prompt them to request access. If management approves, the license upgrade is provisioned automatically, and the employee returns to the tool with the expanded capabilities already active. That experience is organic from the employee’s perspective and trackable from the IT team’s.

Why the Digital Experience Layer Changes What AI Governance Can Accomplish

Satterwhite made a point during the event that often gets left out of AI governance conversations: adoption problems are not always policy problems. Sometimes, adoption roadblocks stem from performance problems. If employees consistently experience frustrating technical issues when navigating to an approved AI tool, such as slow load times, unresponsive features, or device compatibility issues, they will stop using it. That behavioral pattern will look identical in usage data to intentional avoidance, and the governance response will be misdirected.

The broader point here is that AI adoption visibility is not a standalone capability. When it sits within a platform that also understands device health, software performance, employee feedback, and workflow automation, the signal it produces is richer, and the responses available to IT and HR leaders are more targeted. Knowing that a department is underusing Copilot is the starting point. Understanding why—whether it is policy confusion, technical friction, poor onboarding, or a tool mismatch—allows the organization to actually do something about it.

FAQ: Enterprise AI Adoption Visibility

Why do most enterprise AI investments fail to deliver measurable ROI? The most common failure mode is an adoption gap: tools are purchased and deployed, but employees either do not know they are available, do not understand how to use them effectively, or prefer other tools. Without visibility into actual usage patterns, organizations cannot identify or address the gap.

How do you identify which AI tools employees are actually using? Endpoint-level monitoring through platforms like Nexthink tracks which AI applications employees access, how long they spend in those tools, and how usage patterns vary by department and role. This provides a ground-truth view of actual behavior rather than a picture based solely on license provisioning.

How can organizations reduce spending on unused AI licenses? Software metering capabilities track whether individual users are actively engaging with licensed AI tools. After approximately ninety days of non-use, licenses can be reclaimed and redistributed to employees who need them. Automated workflows can handle the reclamation and allow employees to request access again if their needs change.

How do you enforce AI governance without creating employee friction? Just-in-time notifications delivered within the tools employees are actively using are more effective than static policy documents. Redirecting employees to approved tools when they attempt to access unsanctioned ones turns a blocking action into a constructive nudge.

How do organizations keep employees current with rapidly evolving AI features? In-app guides can be updated quickly when new features are released, delivering contextually relevant training within the tool itself rather than requiring employees to find and consult external documentation. Targeted push notifications can also draw employees’ attention to specific new capabilities.