Why companies struggle to put AI to work

A startup called June recently raised 20 million dollars from high-profile tech investors to fix a specific problem: the nightmare of installing AI in a corporate office. Right now, when a large company wants to add AI to its workflow, they often hire teams of specialized consultants—sometimes called forward-deployed engineers—to spend months manually connecting the new technology to their old, creaky software. June is trying to replace those human consultants with software that does the heavy lifting automatically.

Imagine you are trying to hire an assistant to help organize a library that has never been cataloged. You have boxes of books under desks, duplicates hidden in closets, and index cards that don't match the actual shelves. If you hire a new human assistant, they have to spend weeks just mapping out where everything is before they can do any real work. This is the state of most corporate data. Companies use many different platforms—like those for tracking customers or managing payroll—and their data is often scattered, duplicated, or poorly named. When an AI agent arrives, it has no idea which of the three versions of a client's email address is the correct one to use. June works by scanning a company's internal software to create a map of these bottlenecks. It tells the company exactly what needs to be cleaned up and then generates a step-by-step roadmap for how to hook the AI into their specific digital infrastructure without breaking anything.

The rise of these tools highlights a shift in how we think about AI. For a long time, the focus was entirely on making the brain of the AI smarter. Now, the focus is shifting to the hands of the AI—how it actually interacts with the messy, human-made world of corporate spreadsheets and ancient databases. If businesses can make this integration cheaper and faster, we will likely see AI moving from simple chatbots to assistants that can actually manage complex business tasks. For the rest of us, this means the AI we interact with as customers will soon become more capable and less likely to hit errors, simply because companies are finally learning how to let their AI talk to their internal filing cabinets.