As employees continue to use AI tools that haven’t been approved by their organizations, a practice known as shadow AI, much of the conversation has focused on the security risks. Though security is extremely important, there are other concerns that should be considered.
For example, what happens when employees discover useful AI workflows with these tools? And what if those workflows could benefit an entire team or organization, but they’re stuck with the person who discovered them?
Understanding Why Employees Turn to Unapproved Tools
Employees may turn to unapproved AI tools for a number of reasons. Maybe an existing process takes too long, or they need something their company’s systems don’t offer. They might not even realize there’s already an approved tool that could help.
One of the things that makes AI tools appealing is the ability to explore problems using natural language. Employees can ask questions, try different approaches, and work through problems without necessarily having the technical skills to build a solution themselves. In some cases, this might even help them better understand what’s causing the problem in the first place.
Organization support can also play a role. Some companies haven’t established clear policies around AI tool usage. Even when tools have been approved, employees may not have dedicated time to explore what they can do or how they might improve their work. Instead, experimentation becomes something they squeeze into their schedules, often on their own initiative.
Rather than focusing only on which unauthorized tools employees are using, organizations could benefit from asking what employees are trying to accomplish and whether their existing resources are meeting those needs.
From Unsanctioned Experiments to Shared Capabilities
So, what happens when an employee discovers a workflow that could be useful to others? How does an organization make that capability available without introducing unnecessary risks?
A shared capability is more than a prompt someone figured out on their own. It’s a solution others can reuse with approved access, clear ownership, and enough support to keep it working. That same ease of experimentation that draws people to AI can become a team asset, but only if what’s been discovered is captured and governed.
There are a few ways to get there. The workflow might already be possible on an approved tool, and simply needs to be documented or shared so others know how to use it. It could also be integrated into a system employees already rely on. And in some cases, companies are using vibe coding to create internal tools that address a real gap, turning an individual discovery into something a larger group can use.
Not every useful workflow needs its own application, though, and not every company has the resources to build and maintain one. In most cases, it makes sense to use or integrate existing solutions first, and only build something new when the gap is real and someone will own it.
The challenge is figuring out which approach makes the most sense for the organization and the people who will use it.
Choosing the Right Approach
Before deciding how to make a useful capability available to others, organizations should consider a few questions:
Is this workflow already supported by an approved application? Employees may not be aware of functionality that’s already available.
Could the capability be integrated into an existing system? Consider whether employees can access it without having to use another application.
What resources are needed to implement and maintain the solution? Think about ongoing support, not just initial development.
How will the solution meet security and governance requirements? Consider data access, permissions, reliability, and monitoring.
Will the solution be able to evolve with the organization’s needs? Think about flexibility and potential dependence on third-party providers.
Some discoveries may only benefit a handful of employees, while others could improve how entire teams work. The important thing is recognizing when an idea is worth exploring further and finding a practical way to make it available to others.
By paying attention to what employees are accomplishing with unsanctioned AI tools, organizations may uncover opportunities they hadn’t considered.
Image generated by OpenAI’s ChatGPT Images
