Bret Taylor, co-founder and CEO of the enterprise AI startup Sierra, declares that the traditional era of clicking buttons to interact with software is ending. Speaking at the HumanX conference in San Francisco, Taylor argued that future human-computer interaction will shift from navigating complex interfaces to utilizing natural language. This vision centers on Sierra's recent launch of Ghostwriter, an autonomous tool that builds other agents based on user descriptions. Under this "agent as a service" model, users simply state their needs, prompting Ghostwriter to create, deploy, and manage specialized agents to execute tasks without requiring manual software navigation. Taylor, formerly co-CEO of Salesforce, highlighted that many enterprise tools, such as Workday, are underutilized because learning their intricate systems is time-consuming. He predicts users will soon bypass these interfaces entirely to complete tasks like onboarding or enrollment through simple prompts. To illustrate the speed of this new paradigm, Taylor noted that Sierra deployed a custom agent for retailer Nordstrom in just four weeks using Ghostwriter. Sierra's rapid progress is reflected in its financial milestones; the company achieved a $100 million annual revenue run rate less than 21 months after its founding. In September, Sierra secured a $350 million funding round led by Greenoaks Capital, placing its valuation at $10 billion. Taylor emphasizes that while most companies do not want to build software, they do want efficient solutions to their problems, a gap Ghostwriter aims to fill. Despite the optimism surrounding this fundamental shift in software interaction, industry observers caution that fully autonomous AI agents are not yet a reality. Several technologists and investors point out that current implementations rely heavily on human oversight. Companies like Sierra and legal AI firm Harvey employ forward-deployed engineers who must continuously update and fine-tune customer agents to ensure reliability. These engineers play a critical role in maintaining agent performance, indicating that the transition to completely language-driven, hands-off software is still in its early stages. While the trajectory suggests a future where natural language replaces graphical interfaces, the current reliance on human intervention highlights the complexity of achieving true autonomy in enterprise environments.