The argument for AI agents making traditional software obsolete is getting harder to dismiss - especially when it comes from someone who helped build the software.
Bret Taylor, co-founder of Sierra and former co-CEO of Salesforce, made the case directly: the era of clicking through menus, forms, and dashboards is ending. AI agents - software that takes actions on your behalf rather than just generating text - will handle workflows that currently require you to navigate apps manually. In Taylor's framing, clicking buttons is not a feature of good software design. It's a limitation that AI is about to eliminate.
The Interface Becomes Optional
A traditional software interface assumes you know where to go. You click into a CRM, navigate to a contact record, update a field, save. An AI agent inverts this: you describe what you want, and the software handles the navigation. The interface - buttons, menus, dropdowns, all of it - becomes invisible because you're no longer the one operating it.
This is not a new prediction. Voice assistants launched a decade ago with similar promises. What's different now is that agents are reliable enough to execute multi-step tasks without constant error. Taylor is speaking from direct commercial experience. Sierra builds AI customer service agents that talk to customers, look up account data, and resolve issues without a human navigating any interface at all. He's watched enterprises shift real workflows to conversational AI - not as a pilot, but as their primary operating model.
Who Gets Disrupted First
Enterprise software is the obvious first target. CRMs, ERPs, and project management platforms were designed to route information between people. Agents can handle a significant portion of that routing directly. If a customer contacts support, an agent can query the database, compose a response, and close the ticket without anyone logging into anything.
Consumer software is messier. Photo editors, video tools, and design platforms involve judgment calls that don't translate cleanly to natural language. But even there, the direction is clear: AI is absorbing the mechanical parts of creative workflows and leaving the decisions to the human.
For the people using AI tools daily - marketers scheduling content, analysts pulling reports, support reps handling tickets - the shift is already visible. Copilots embedded in Word, Slack, and Salesforce handle tasks that once required menu navigation. Taylor's argument is that this trend doesn't plateau at "helpful assistant." It keeps going until the interface is optional.
Sierra's own trajectory supports the point. AI-powered customer service was a niche category two years ago. It's now a product segment with serious competition. The companies still routing everything through traditional UIs are competing against companies that have skipped the UI entirely.