Marketers Aren't Just Adopting Agents. They're Hooked.
The one thing you need to know in AI today | AI Ready CMO
A joint study from Perplexity and Harvard just gave us the first large-scale behavioral data on how people actually use AI agents. Let’s pat ourselves in the back: marketers are leading the charge.
The research (summary, full PDF) analyzed hundreds of millions of anonymized interactions from Comet, Perplexity’s AI-powered browser, between July and October 2025. Unlike chatbots that just answer questions, Comet actually takes control of your browser—editing docs, sending emails, navigating marketplaces, posting to social. It’s “agentic” in the truest sense: you give it a goal, it executes the steps.
Marketing and sales professionals aren’t just early adopters. They’re power users, scoring above tech workers.
The study uses something called “Agent Usage Ratio” (AUR) to measure how intensively different groups use the agent relative to their population size. Marketing & Sales scored 1.46—the highest of any occupation cluster, including tech workers. Digital marketing and social media professionals in particular hit 1.51.
That means once marketers adopt an agent, they don’t just experiment; they embed it into their daily workflow at a rate that outpaces every other profession.
The top environments tell the story: LinkedIn, Google Docs, Instagram, X, email services. This isn’t “write me a tagline”—it’s prospecting, content scheduling, document drafting, and inbox management, all executed by the agent inside the actual platforms.
Business development and sales roles follow the same pattern. LinkedIn dominates—over 90% of professional networking queries happen there—with tasks like filtering profiles, sending connection requests, and drafting outreach at scale. The agent becomes, effectively, a LinkedIn operator. Sales work has always involved countless micro-actions: search, filter, qualify, message, follow up. Browser agents are a natural fit precisely because they can execute those steps across web interfaces without you toggling between seventeen tabs.
Creative professionals, meanwhile, are the counterintuitive finding.
Despite all the hype about AI and creativity, the Arts, Entertainment, & Design cluster actually under-indexes in both adoption and usage intensity.
When they do use agents, it’s heavily concentrated in Canva, YouTube, and Google Docs—more “creative-adjacent” research and organization than hands-on generation. The plausible read: creative work still relies on taste, iteration, and fine-grained feedback loops that a slightly-off agent can disrupt more than help. Or maybe it’s that much of creative work still happens in desktop-native tools (Adobe apps) that aren’t purely browser-mediated. Either way, if you’re running a creative team and assuming agents will immediately reshape your workflows, the behavioral data suggests slower adoption than you might expect.
And one more interesting data point from a leadership perspective: the gap between early adopters and everyone else is massive.
A user from the first cohort (July) makes roughly nine times as many agentic queries as someone who joined at general availability in October. That’s not a rounding error.
A small group of marketers is redesigning their entire workflow around agent delegation while the majority is still experimenting. If your mental model of “how marketers use AI agents” is based on what you see from the vocal early adopters, you’re probably overfitting to the exception. (We are guilty as charged.)
The real implication here isn’t about adopting agents faster. It’s about understanding which tasks are worth delegating and which require your judgment. The study explicitly calls this out as the open research question: optimal human-agent collaboration design, especially for high-stakes tasks.
For now, marketing and sales have found their sweet spot—high-volume, low-irreversibility actions executed across browser surfaces. The question for leadership is whether you’re being deliberate about where you draw that line, or just waiting for your team to figure it out on their own.
3 AI Marketing Tools To Try Today
Quillbot
Rewrite and refine content without starting from scratch—Quillbot paraphrases, summarizes, and improves grammar instantly for cleaner messaging.
CloudTalk
Transform customer calls into actionable intelligence—CloudTalk transcribes conversations, analyzes sentiment, and surfaces insights to improve sales and support.
Podcastle
Record, edit, and enhance podcast audio like a pro—Podcastle removes background noise, balances levels, and generates transcripts without technical skills.