AI Browser Agents Radically Transform How We Use The Internet

Most of us were introduced to the internet through web browsers; who remembers Netscape Navigator? Or the more recently departed Internet Explorer?

For 30 or so years, they’ve been an essential everyday business tool, and during that time, they haven’t changed much.

Until now, that is. AI is upending everything, and that certainly applies to the way we search for information online.

We saw the first stage of this transformation with the introduction of generative search. And while businesses and the rest of us are still coming to terms with the implications of that, here comes the next big thing: browser agents.

This is the term for a new generation of web browsers that, rather than simply helping us find the information we need to complete tasks, will take action to get them done.

This means that instead of searching, clicking links and scrolling through content, we will delegate tasks to the browsers themselves. The idea is that they can see what we see and can do, or at least try to do, anything we can do.

This has the potential to be a significant shift in how we use the internet. But the technology is at an early stage and comes with some pretty serious risks attached.

So, let’s take a look at what agentic browsing could mean for the future of the internet, information, and the way humans work with machines.

How Does Agentic Browsing Work?

A number of agent-enabled web browsers have been released to the public in recent months: these include ChatGPT Atlas (currently Mac-only), Opera Neon and Comet from Perplexity.

The most obvious difference between these and traditional browsers is that they have a natural-language, chatbot-style interface built in. When you start up the browser, this is a very prominent feature of the UI.

But rather than simply answering questions or generating information like ChatGPT, the browser can translate your written instructions into actions; clicking links, filling out forms, visiting sites and operating web apps.

This is thanks to computer vision, which lets the AI “see” what’s on the screen, parsing the language and visual information to learn how to operate web pages and apps.

It’s important to remember that this new generation of browsers is distinct from, and a step ahead of, AI-enabled browsers, like new versions of Google’s Chrome. AI-enabled browsers have some AI features built in, such as built-in generative AI search or the ability to summarize websites, but stop short of being able to actually complete tasks for us.

The potential uses for agentic browsers are practically unlimited; in theory, they can do anything we can do with a web browser. However, in reality, the technology is very new, highly experimental, and, in a lot of cases, slower and less capable than we are.

But let’s take a look at some of what is possible today (to varying degrees of success):

So What Can They Do?

You can use a browser agent as a research aid, asking it to visit various websites, pulling information and generating a summarized report that answers critical questions, such as “help me compare SaaS CRM vendors that can automate customer relations for my business.”

It could also help make travel arrangements, and in theory go as far as making bookings (although there are reasons you probably don’t want to let it do this just yet, more on that below): “Book me an economy-class round-trip flight and a hotel in New York, close to the conference center venue. The hotel must have a swimming pool.”

It can also be an administrative assistant for work: “Summarize all of the open tasks on my project planner, create a prioritized action plan with next steps and add reminders to my calendar for important deadlines.”

Or it can help out with shopping tasks: “Place an order for an ice-cream maker. It must be able to make one liter of ice cream, come in blue, and the vendor must offer free returns.”

As you can see, this means it combines the ability of generative search to reduce manual research and information retrieval, with the ability of AI agents to take actions towards accomplishing goals.

But Wait…

The arrival of agentic browsers clearly has the potential to radically change the way we use the internet.

Web browsers are evolving from passive gateways for accessing information into automated productivity engines. Rather than wasting time clicking through search results and multiple tabs, we will focus on more valuable work like creativity and high-level decision-making.

For businesses, this means rethinking how websites and apps connect with customers from the ground up, as their audience on these channels increasingly includes intelligent machines.

However, it also comes with some very important caveats:

Firstly, the technology is at a very early stage; even basic tasks like browsing, opening web pages, and scanning for information can be slow and inaccurate, and the information agents think is important won’t always align with our own priorities.

More worryingly, though, are the security concerns. All of the agentic browsers mentioned here can, in theory, be given access to everything from your email to your online banking. But there’s little clarity about how information is protected, so I certainly wouldn’t recommend doing this just yet.

And if we’re trusting browser agents to make buying decisions, how do we know they’ll spend our money wisely? With the limited context available to them, they could easily make choices that, while seemingly logical, miss the nuances of our human preferences and intentions.

These concerns are enough to ensure that, for now, browser agents will remain an exciting experimental technology.

It’s worth remembering, though, that these were all things we worried about when the first web browsers became available, three decades ago. Back then, many people rightly questioned the wisdom of putting personal information, particularly financial information, into the new thing called “the internet”. But today, most people shop and manage their finances online in relative safety.

Back in the day, the evolution of the internet from a hobbyist playground for computer enthusiasts, to an essential utility for business and everyday life took careful planning, orchestration and legislation.

The emergence of this new, agentic internet is likely to follow a similar, if accelerated, path. Today, however, the appearance of this first generation of agentic browsers offers a fascinating glimpse into the evolving relationship between humans and machines.