Imagine navigating the digital world entirely through sound. That's the reality for millions of blind computer users who rely on screen readers. These tools convert on-screen text and interface elements into spoken words, but the experience can be clunky and inconsistent. Different applications have different keyboard shortcuts and layouts. What works in Microsoft Word might not work in Excel, creating a steep learning curve and slowing down productivity. What if there was a way to interact with any software using the same intuitive commands? Researchers are exploring how large language models (LLMs), the technology behind AI assistants like ChatGPT, could create a more uniform experience. They’ve developed a system called Savant that lets users control applications with natural language. Instead of memorizing complex key combinations, a user could simply say “set the margin to narrow” or “insert a pie chart.” Savant interprets the command, figures out the corresponding actions within the specific application, and executes them automatically. In a user study with blind participants, Savant significantly improved efficiency and usability. Participants performed tasks faster, with fewer keystrokes, and reported higher satisfaction. While promising, there are still challenges. Savant currently struggles with complex commands, like those involving multiple steps or different applications, which researchers are hoping to improve in the future. Making it work seamlessly with pop-up windows and sub-menus is also on their roadmap. If successful, AI-powered tools like Savant could bridge the gap between complex software and accessibility, empowering blind users with a smoother, more intuitive computing experience.