Aravind Srinivas, the CEO of the ambitious AI startup Perplexity, has a clear and startling vision for the future of work. It begins with a simple prompt and ends with the automation of entire professional roles.
“A recruiter’s work worth one week is just one prompt: sourcing and reach outs,” Srinivas stated in a recent interview with The Verge’s Decoder” podcast, a prediction that serves as both a mission statement for his new AI-powered browser, Comet, and a stark warning for the modern knowledge worker.
His company is at the forefront of a new technological arms race to build not just a smarter search engine, but a true AI agent. Think of it as a digital entity capable of carrying out complex, multi-step tasks from start to finish. According to Srinivas, the most natural place for this revolution to begin is the one tool every office worker already uses: the web browser. And the first jobs in its sights are those of recruiters and executive assistants.
The Automation of Expertise
For years, the promise of AI has been to assist, not replace. But the vision Srinivas lays out is one of replacement by a vastly more capable assistant. He describes an AI agent as something that can “carry out any workflow end to end, from instruction to actual completion of the task.”
He details exactly how Comet is being designed to absorb the core functions of a recruiter. The agent can be tasked to find a list of all engineers who studied at Stanford and previously worked at Anthropic, port that list to a Google Sheet with their LinkedIn URLs, find their contact information, and then “bulk draft personalized cold emails to each of them to reach out to for a coffee chat.”
The same logic applies to the work of an executive assistant. By having secure, client-side access to a user’s logged-in applications like Gmail and Google Calendar, the agent can take over the tedious back-and-forth of scheduling. “If some people respond,” Srinivas explains, the agent can “go and update the Google Sheets, mark the status as responded or in progress and follow up with those candidates, sync with my Google calendar, and then resolve conflicts and schedule a chat, and then push me a brief ahead of the meeting.”
This is a fundamental re-imagining of productivity, where the human role shifts from performing tasks to simply defining their outcomes.
A Six-Month Horizon
While Comet cannot execute these most complex, “long-horizon” tasks perfectly today, Srinivas is betting that the final barriers are about to fall. He is pinning his timeline on the imminent arrival of the next generation of powerful AI.
“I’m betting on progress in reasoning models to get us there,” he says, referencing upcoming models like GPT-5 or Claude 4.5. He believes these new AI brains will provide the final push needed to make seamless, end-to-end automation a reality.
His timeline is aggressive and should be a wake-up call for anyone in these professions. “I’m pretty sure six months to a year from now, it can do the entire thing,” he predicts. This suggests that the disruption isn’t a far-off abstract concept but an impending reality that could reshape entire departments before the end of next year.
From Browser to OS: A New Layer of Automation
Srinivas’s ambition extends far beyond building a better browser. He envisions a future where this tool evolves into something much more integral to our digital lives.
“That’s the extent to which we have an ambition to make the browser into something that feels more like an OS where these are processes that are running all the time,” he says.
In this new paradigm, the browser is no longer a passive window to the internet but an active, intelligent layer that manages your work in the background. Users could “launch a bunch of Comet assistant jobs” and then, as Srinivas puts it, spend their time on other things while the AI works. This transforms the very nature of office work from a series of active inputs to a process of delegation and oversight.
Liberation or Mass Displacement?
What happens to the human worker when their job functions are condensed into a single prompt? Srinivas offers an optimistic view, suggesting that this newfound efficiency will free up humanity’s time and attention. He believes people will spend more time on leisure and personal enrichment, that they will “choose to spend it on entertainment more than intellectual work.” In his vision, AI does the drudgery, and we get more time to “chill and scroll through X or whatever social media they like.”
But this utopian view sidesteps the more immediate and painful economic question: What happens to the millions of people whose livelihoods are built on performing the very tasks these agents are designed to automate? While some may be elevated to the role of “AI orchestrator,” many could face displacement.
The AI agent, as described by one of its chief architects, is not merely a new feature. It is a catalyst for a profound and potentially brutal transformation of the white-collar workforce. The future of work is being written in code, and according to Srinivas, the first draft will be ready far sooner than most of us think.
