Bret Taylor, co-founder and CEO of the enterprise AI agent company Sierra, said in an interview published Thursday that the era of clicking buttons in software is over. The claim is not entirely new, but the person making it and the company behind him give it more weight than it might otherwise carry. Sierra reached $100 million in annual recurring revenue in under 21 months from founding, was last valued at $10 billion after a $350 million funding round, and counts Taylor, a former co-CEO of Salesforce and current chair of OpenAI's board, among its leadership. When someone with that background says that the fundamental paradigm of software interaction is being replaced, it is worth examining what they actually mean and where the evidence points.
Taylor's thesis is that users will increasingly accomplish tasks by describing what they want in natural language rather than navigating interfaces. The traditional software model, selecting options from menus, clicking through screens, filling in forms, is optimised for a world in which computers need explicit instruction at every step. AI agents change the instruction model: instead of the user directing each action, the user states an objective and the agent determines how to accomplish it. Sierra's product, Ghostwriter, is designed to let companies deploy agents that can handle customer interactions end to end without a human representative or a conventional web interface in the loop.
The enterprise traction is real. Sierra's customer base includes companies deploying agents for customer service at scale, and reaching $100 million ARR faster than most enterprise software companies means someone is paying for it and finding value. The $10 billion valuation, while large, reflects a market willing to bet that the agent paradigm will take a significant share of the enterprise software space that currently belongs to conventional SaaS tools.
The more interesting tension in the story is a gap between the vision and the current reality of deployment. Despite Sierra's framing of agents as autonomous, Sierra itself employs what are called forward-deployed engineers: specialists who work at customer sites to tune and update agents on an ongoing basis. The gap between what agents can do in a demo and what they reliably do in production, across the full range of edge cases that real enterprise workflows throw at them, is still significant enough that human oversight remains a practical requirement. Taylor's "era of buttons is over" claim is most accurately read as a directional statement about where things are heading rather than a description of how Sierra's product works today.
That gap matters for evaluating the broader claim. The transition from click-based interfaces to agent-based ones will not be sudden or clean. It is more likely to be gradual, with agents taking over well-defined, high-volume tasks first, while complex or sensitive interactions retain human involvement or conventional interfaces for longer. The process is already underway in customer service: basic query resolution, order status, return initiation. Whether it extends to the full range of software interaction that Taylor is describing is a longer-horizon question.
Taylor's position as both Sierra's CEO and OpenAI's board chair makes him one of the more connected individuals in the current AI industry structure. His public statements on the agent transition carry both genuine insight into where leading labs are headed and an obvious commercial interest in accelerating the adoption of exactly the kind of agent products Sierra sells. Both things can be true at once. The useful way to read his "buttons are over" pronouncement is less as a factual claim about current software and more as a signal of how the people building the next generation of enterprise tools are thinking about where value is going to move in the next several years.