Phantom Of The Opera: AI Agent Now Lurks Within Browser, For The Lazy

The Opera web browser now boasts "agentic AI," meaning users can ask an onboard AI model to perform tasks that require a series of in-browser actions.

It's basically delegating one's web browsing to a personal assistant who's not particularly competent but can manage some things. Fun as a technology demonstration perhaps, but c'mon.

The AI agent, referred to as the Browser Operator, can, for example, find 12 pairs of men's size 10 Nike socks that you can buy. This is demonstrated in an Opera-made video of the process, running intermittently at 6x time, which shows the user has to type out the request for the undergarments rather than click around some webpages.

The AI, in the given example, works its way through eight steps in its browser chat sidebar, clicking and navigating on your behalf in the web display pane, to arrive at a Walmart checkout page with two six-packs of socks added to the user's shopping cart, ready for payment.

You could do so, too, in 15 seconds or so by visiting Walmart's online store and clicking a few times. But then your AI agent would be out of job, ironically enough.

Other tasks such as finding specific concert tickets and booking flight tickets from Oslo to Newcastle are also depicted, accelerated at times from 4x to 10x, with the user left to authorize the actual purchase.

Browser Operator runs more slowly than shown in the video, though that's actually helpful for a semi-capable assistant. A more casual pace allows the user to intervene at any point and take over.

"For more than 30 years, the browser gave you access to the web, but it has never been able to get stuff done for you," Krystian Kolondra, EVP of Opera Software, argued in a statement somewhat overlooking the billions of CPU cycles needed to parse a modern HTML page with CSS and JavaScript.

"Now it can. This is different from anything we’ve seen or shipped so far. The Browser Operator we’re presenting today marks the first step towards shifting the role of the browser from a display engine to an application that is agentic and performs tasks for its users."

The phrase "now it can" actually means "now it can for some" – the Browser Operator is a preview feature that some users may have access to but isn't yet generally available. The latest macOS builds for Opera and Opera Developer did not yet support Browser Operator when we hunted around the Settings and Flags menus to enable it.

According to the biz, Browser Operator will debut via the firm's AI feature drop program in the near future.

• AI agents? Yes, let's automate all sorts of things that don't actually need it
• Workday talks up AI agents platform that will reap rewards of staff cuts
• We can clone you wholesale: Boffins build ML agents that respond like specific people
• AI agent promotes itself to sysadmin, trashes boot sequence
• Kelsey Hightower on dodging AI and the need for a glossary of IT terms

Browser Operator runs locally, according to Opera. "It protects user privacy by not relying on screenshots or video capture of the browsing session, nor on a version of the browser running in the cloud or a virtual machine," the developer claims.

Those familiar with the Windows Recall rollout debacle may appreciate the lack of screen captures.

Opera committed to mingling AI with the web in 2023 when it added its Aria AI Assistant to its browser. To date, the verdict on social media and in Opera forums appears to be mixed.

Its Aria settings page includes this warning: "AI Services should not be employed for any legal, medical, or advisory purposes. We advise you to apply caution and check facts."

We advise you to apply caution and check facts

OpenAI and Anthropic have both released agent-oriented services, specifically OpenAI Operator and Anthropic's computer use API for Claude 3.7 and 3.5 Sonnet.

Perhaps surprisingly, the Wall Street Journal in January reported that some companies are actually using AI agents.

Gartner last week forecast that within a few years AI agents will play a significant role in business operations. "By 2028, Gartner predicts that 33 percent of enterprise software applications will include agentic AI, up from less than 1 percent in 2024, with at least 15 percent of day-to-day work decisions being made autonomously through AI agents," the IT consultancy said.

Keep in mind that Gartner has been wrong before. In 2014, Gartner research director Peter Sondergaard predicted, "one in three jobs will be converted to software, robots and smart machines by 2025." Only three years later, Gartner said AI will create more jobs than it destroys. ®