The excitement surrounding artificial intelligence is undergoing a massive shift. For the past two years, the industry focused almost exclusively on large language models that could write poems, summarize emails, and chat with users. Now, Perplexity CEO Aravind Srinivas argues that the “chatbot era” is reaching its limit. During a wide-ranging discussion about the future of the tech industry, Srinivas explained that the next major surge in value will come from “agentic” systems—AI that doesn’t just talk, but actually performs complex tasks to solve real-world problems.
According to Srinivas, the early excitement for AI was driven by a novelty factor. People were amazed that computers could generate human-like text. However, we are moving past that stage very quickly. Businesses and consumers are demanding more than just a smart conversation; they want a digital worker that can execute workflows. He believes that the companies capable of building agents that manage logistics, coordinate software environments, and make autonomous decisions will define the next $1 billion success stories in the sector.
The current valuations of many AI startups appear to be disconnected from their actual revenue-generating capabilities. Investors have poured billions into companies that only offer a slightly better interface for public language models. Srinivas warned that the market will soon undergo a “reality check.” Investors will eventually stop rewarding simple chatbots and start demanding proof of “agentic” efficiency. If a company cannot show how its software reduces costs or increases output by at least 1.5% to 5% across an enterprise, it will struggle to justify its high price tag in the coming years.
The push toward agentic AI is not just a trend for startups; it is a defensive strategy for every major tech player. When a bot can act as a personal assistant, a researcher, or a coding companion, it creates a “stickiness” that is difficult for competitors to break. This is why we see firms racing to integrate these agents into browsers, operating systems, and office suites. If an AI agent can handle your entire travel booking process—from flight selection to hotel check-in—you are much less likely to switch to a platform that requires you to do the heavy lifting yourself.
This shift creates a massive opportunity for platforms like Perplexity, which aim to provide reliable, source-backed information rather than just generative summaries. If an AI agent is going to make decisions on your behalf, it needs a foundation of truth. If the model relies on “hallucinations” or incorrect facts, the agentic workflow will collapse. Srinivas believes his company’s focus on accuracy makes it the perfect partner for the next generation of autonomous business agents.
Of course, this transition to agentic AI introduces significant security risks. If you give a piece of software the permission to book your flights, send your emails, or access your bank details, you are effectively handing over the keys to your digital life. Security experts warn that companies must build “sandboxes” where these agents operate, ensuring they cannot access sensitive data without explicit user verification. The industry is currently scrambling to build these security layers, knowing that a single high-profile exploit could destroy consumer trust in the entire AI agent ecosystem.
The economic impact of this change could be profound. Estimates suggest that if automated agents successfully handle just 10 percent of routine office tasks, it could free up millions of hours of labor every year. This productivity boost is what venture capitalists are really betting on. They are looking for the “office in a box” solution, where a single subscription to an AI agent can replace a suite of expensive, manual software tools. It is a fundamental disruption to the B2B software market, and it explains why funding rounds for agent-focused startups remain so hot.
Musk’s recent moves with SpaceXAI and other companies show that he agrees with this sentiment. He is moving his entire AI strategy toward agents that can operate robots and interact with the physical world. This is the logical end-game for the industry. A chatbot is just a toy; an agent that can interact with the physical and digital world is a utility. Companies that understand this distinction will lead the market, while those that remain stuck on “chat-first” designs will likely fall behind.
As we look toward the end of 2026, the question for every major tech firm is no longer “is your AI smart?” but rather “is your AI useful?” The era of the “AI search engine” is fading, replaced by the era of the “AI executor.” We are moving toward a world where your laptop and phone will be populated by tiny, specialized agents, each handling a specific part of your day.
The transition won’t be easy. There are huge engineering hurdles to clear, from reducing the latency of these agents to ensuring they don’t break when the internet connection flickers. But the path is clear. The next $1 billion unicorn won’t be the company with the best poem-writing bot; it will be the one that builds the best agent for getting things done. The AI market is growing up, and it is leaving the chatbots behind.
