What is the real value of AI Agents?
2025 was supposed to be the year of AI agents. You’ve probably heard the pitch: autonomous assistants that book your travel, manage your calendar, and run your business while you sleep.
But here’s the thing: until now, we’ve had a lot of hype and very little data. We didn’t actually know how people were using AI agents in the real world.
Well, we have a new study to help answer this question around AI Agents.
Harvard researchers teamed up with Perplexity AI to publish the first large-scale field study of AI agent adoption. They analysed hundreds of millions of user interactions. And what they found challenges a lot of the assumptions about what AI agents are actually good for.
So today I want to break down what this study tells us, and why it matters for how you should be thinking about AI in 2026.
The Study
First, let’s be clear about what we’re talking about.
AI agents are different from chatbots. A chatbot answers questions. An agent takes action. It can navigate websites, fill in forms, create documents, manage your email, it operates in the real world on your behalf.
Perplexity launched an AI-powered browser called Comet in July 2025, with an integrated assistant. Harvard researchers analysed how people actually used it, hundreds of millions of queries from real users doing real work.
They asked three questions: Who’s adopting AI agents? How intensively are they using them? And what are they using them for?
The data
1. It’s About Cognitive Work, Not Convenience
Here’s the headline finding that surprised me.
The popular narrative is that AI agents are digital concierges; they’ll book your hotel, order your groceries, and handle the boring stuff. And yes, they can do that.
But that’s not what people are actually using them for.
57% of all agent activity is cognitive work. Productivity and workflow tasks account for 36% of queries, things like creating documents, managing emails, and summarising information. Learning and research account for another 21%, students using agents to work through course materials, and professionals researching topics for work.
People aren’t just offloading chores. They’re using agents to think better, learn faster, and produce higher-quality work.
2. Knowledge Workers Are Pulling Ahead
The study found clear patterns in who’s adopting AI agents.
Over 70% of users work in knowledge-intensive fields such as digital technology, academia, finance, marketing, and entrepreneurship. Which fields have the fewest AI agent users? Jobs that involve physical environments, energy, agriculture, and construction.
Adoption correlates strongly with GDP per capita and education levels. Wealthier, more educated countries are adopting faster.
This isn’t just interesting trivia. It’s a signal. If you work in a knowledge-intensive field and most of my listeners do, your competitors and peers are already building AI agents into their workflows. The gap is widening.
3. Usage Evolves And Sticks
This is the finding that tells us where this is going.
New users tend to start with low-stakes queries, travel, entertainment, and general questions. They’re testing the water.
But over time, usage shifts. People move toward productivity, learning, and career-related tasks. And once they start using agents for cognitive work, once they’ve used an agent to summarise a financial report or debug a piece of code, they rarely go back to simpler use cases.
The researchers compared it to the early days of personal computers. PCs were sold as tools for recipes and games. They became indispensable because of spreadsheets and word processing. AI agents are following the same trajectory.
Looking ahead to 2026
So what do you do with this?
If 2024 was about experimenting with chatbots, and 2025 was about integrating AI into specific tasks, 2026 is about agentic AI, AI that doesn’t just answer questions but takes action on your behalf.
This study gives us actual evidence that the value is real. People who adopt AI agents use them for serious work, and they stick with them.
My suggestion: if you haven’t tried an AI agent yet, start small. Use one for a productivity task, summarising a long document, drafting an email, or researching a topic. See how it changes your workflow.
Because the data is clear. The early adopters aren’t using AI agents to book hotels. They’re using them to think. And that’s a competitive advantage that compounds over time.
This Harvard and Perplexity study is the first real evidence base for AI agent adoption. It tells us that the value isn’t in convenience, it’s in cognitive leverage.
If you’re in a knowledge-intensive field, 2026 is the year to get serious about agentic AI.
Know it. Use it. Govern it.
About the Podcast Host
Joanne Sweeney is a specialist in public-sector marketing and digital communications. She’s the author of Public Sector Marketing Pro and the creator of the Professional Diploma in AI for Public Sector Marketing. With years of experience in training, consulting, and advocating for digital transformation, Joanne provides insight and guidance for government agencies navigating the digital landscape. She is also the host of the award-winning Public Sector Marketing Show podcast.
