Your employees aren’t just using software. They’re building it.

This is already happening inside organizations: employees are using AI to build useful tools the moment they encounter a need. Instead of passing a request to a development team or submitting a request to purchase new software and waiting for a solution, they open a chat, describe what they need, and build a solution themselves. It’s a major leap in employee autonomy and speed, but it also creates a new challenge for organizations: how do you let employees keep building quickly without losing control over applications, access, and data?

As a product manager in the software industry, a big part of my job was taking a very specific customer need and looking for the broader problem behind it. That is almost an inherent requirement of commercial software: to justify development, a need has to be shared by enough customers. Specific needs within a single team or department are often left unaddressed.

Those needs, of course, haven’t disappeared. Every organization has hundreds of them: a tool tailored to a specific onboarding process, a pricing calculator for the sales team, or an interface that connects several systems for the finance department. Until now, most of these needs were simply too narrow and specific to justify building custom software.

Chat has already become a place where employees go to solve problems. Now that they can also build applications there using natural language, employees far beyond traditional development teams can build for themselves. They describe what they need in a chat and create the tool they were missing.

This isn’t the first time a capability once reserved primarily for professionals has become accessible to every employee. PowerPoint, for example, didn’t invent the presentation. Before it, producing a professional presentation often required a graphics department or an external provider to create the slides. PowerPoint enabled employees to do it themselves, turning presentation creation into a routine part of office work.

But when employees start building software, a new organizational challenge emerges. Building is only the beginning. What happens when an employee wants ten or a hundred other people to use what they built? Where does the application run? Who is allowed to access it? Which systems and data can it access? Who pays for the AI usage it consumes? Who makes sure it complies with the organization’s security requirements and policies? And what happens when the employee who built it changes roles or leaves the company?

At the same time, the major AI providers, which already enable users to build applications, are beginning to extend their platforms to running and managing them as well. As more enterprise software lives inside these platforms, organizations become increasingly dependent on them. Companies will need to think not only about who can build software, but also where that software lives, who owns it, and who controls it over time.

The next layer of enterprise software won’t come only from software vendors or development teams. It is already beginning to emerge from within the organization itself. The challenge will be enabling employees to build solutions tailored to their organization without losing control over them.