Perplexity CEO Aravind Srinivas at HubSpot INBOUND 25Ron Schmelzer at HubSpot INBOUND 25
Talking at HubSpot’s INBOUND 2025 convention, Aravind Srinivas, CEO of Perplexity, shared a message that cuts towards right this moment’s noisy AI hype cycle. Whereas most AI firms race to make fashions extra all-encompassing and humanlike, he argued the actual breakthrough lies in spurring curiosity.
“Earlier than AI, should you had been interested by one thing, you needed to collect a workforce, rent consultants, debate in a committee,” he stated. “Now you may simply ask.” For Srinivas, the worth of AI isn’t in changing work, however in serving to folks ask higher questions. Essentially the most profitable staff, he claims, would be the ones who enter conferences with sharper questions, not pre-baked solutions.
Curiosity as Leverage
Srinivas in contrast curiosity to leverage in enterprise and science. He pointed to moments in historical past the place curiosity drove innovation. Transistors grew out of challenges to the boundaries of vacuum tubes. John Deere reshaped agriculture by asking if metal might change brittle iron.
“In case you aren’t listening to ‘that’s not possible’ or ‘why would you even ask that,’” he stated, “you’re in all probability not asking onerous sufficient questions.”
This attitude echoes feedback from Satya Nadella, Microsoft’s CEO, who has emphasised that AI ought to assist folks “purpose over knowledge, not drown in it.” The place Nadella leans towards reasoning instruments for enterprise scale, Srinivas takes a extra private strategy, specializing in how people can remodel the best way they suppose inside a corporation.
Why Accuracy Issues When AI Goes Unsuitable
Perplexity launched in December 2022, only a week after ChatGPT’s massive launch. On the time, customers discovered AI’s errors entertaining, sharing screenshots of absurd outputs that went viral. Perplexity’s Buyers informed Srinivas that specializing in a drier strategy, including citations and verifications to responses,made solutions boring. He disagreed, “Solely an correct reply results in the subsequent good query.”
That view has develop into Perplexity’s calling card. Its product doesn’t simply give responses. It cites sources and nudges customers towards follow-up questions. “Curiosity doesn’t cease with a solution,” Srinivas stated. “It begins there.”
Critics of LLMs level out that hallucinations stay a persistent downside. Gary Marcus, a cognitive scientist and long-time AI critic, has argued that belief in AI hinges on verifiability, not fluency. In that sense, Perplexity’s mannequin, grounded responses with seen sources, aligns extra together with his name for reliability than with Silicon Valley’s choice for type, transferring quick, and breaking issues.
Constructing Assistants That Really Assist
Srinivas desires greater than citations. His workforce is constructing Comet, a browser assistant that watches Slack, electronic mail, and dashboards, then brings the proper context into view. As a substitute of copying and pasting right into a chatbot, the instrument surfaces materials as you’re employed. He calls it a second mind, all the time close by, by no means in the best way.
Different startups are chasing related concepts. Adept is coaching AI to click on round inside software program. Rewind desires to report each display interplay for recall later. OpenAI has experimented with customized GPTs that deal with slender duties. Perplexity takes a special tack: one assistant that adapts to you, not the opposite manner round.
“You shouldn’t want immediate engineering lessons to get your job completed,” Srinivas stated.
Competing Philosophies within the AI Race
AI firms are drifting aside in how they see the longer term. Some race to construct larger, quicker fashions with broad capabilities. Perplexity is betting that reliability will win over flashy scale. Anthropic has tried to separate the distinction with Claude, designing it to say nothing when it isn’t positive. Critics say that makes it timid.
Ethan Mollick, a Wharton professor finding out AI in enterprise, places it bluntly, “For college students, fluency issues. For executives, reliability does.” That explains why client chatbots, vulnerable to hallucinations, catch hearth on social media however stumble in boardrooms.
Srinivas didn’t sugarcoat it. “Watch out for AIs that all the time inform you what you wish to hear. These aren’t assistants. These are sycophants.”
Constructing Curiosity as a Tradition
Behind the product pitch, Srinivas saved returning to a cultural theme. He argued that conferences ought to be judged much less by polished slides and extra by the standard of the questions raised. Too typically, forms, standing updates, and formatting bury curiosity. A great assistant, in his view, would clear the busywork so folks can get again to the joys of inquiry.
That concept borrows from Slack’s co-founder Stewart Butterfield, who as soon as stated software program ought to cut back “work about work.” Srinivas extends it. He thinks AI can revive the artistic spark inside organizations, giving staff extra room to suppose and query as a substitute of simply executing operations.
The onerous half will probably be execution. Digital assistants have been promised earlier than, and Siri, Alexa, and Google Assistant all stumbled. Perplexity is wagering {that a} deal with accuracy and curiosity will preserve it from the identical destiny. Whether or not that pays off depends upon whether or not folks really need greater than fast solutions from LLMs.
