Ad industry foresees another publisher traffic hit as OpenAI’s AI agent, Operator, arrives

AI agents like Operator can be deployed to autonomously perform specific tasks or achieve designated goals, independently collecting data and interacting with their environments to take action.

Unlike some existing AI tools that require user intervention for execution, Operator can independently handle tasks like booking travel and ordering groceries.

“You give it a task and it will execute it,” OpenAI wrote in a blog post today. “Using its own browser, it can look at a webpage and interact with it by typing, clicking, and scrolling.”

The program is currently available as a research preview for Pro users in the US. The company says it will make adjustments to Operator based on user feedback.

OpenAI CEO Sam Altman has previously signaled his bullishness on agentic AI, writing in a blog post on January 5 that “we may see the first AI agents … materially change the output of companies” in 2025.

Operator, plans for which were first announced in November, is designed to serve as a general-purpose agent, able to autonomously complete multi-step tasks in a web browser.

In this sense, the program appears to be most akin to Anthropic’s agentic Claude Computer Use, rolled out in October, which can take control of a web browser autonomously to complete tasks like planning a vacation itinerary. Last month, Google also unveiled a so-called ‘computer-use agent’ dubbed Mariner.

But AI agents aren’t just made for planning trips and booking appointments. They’re increasingly finding applications in the workplace – which could transform how organizations conduct business and manage operations.

As Microsoft’s vice-president of telecommunications, media, entertainment and gaming Silvia Candiani said in an exclusive interview with The Drum at CES 2025 earlier this month, “You could have an AI agent doing some of the back office work, or you could have an AI agent specific to media programming, or an AI agent for content creation. So you can have [a whole range of] specialized tools that you can govern…”

The market is already seeing an influx of AI agent solutions designed to cater to enterprise-level applications. Just this month, Oracle launched AI agents for sales professionals and Accenture introduced 12 industry-specific agent solutions. In September of last year, Salesforce rolled out its enterprise AI agent platform Agentforce; CEO Marc Benioff has said that he hopes to see 1bn Agentforce agents deployed by the end of 2025. Two months after that announcement, Microsoft debuted Copilot Studio, which enables teams to build autonomous IT agents. Oracle, SAP and Nvidia have also joined the race.

Considering that Operator is meant to be a general-purpose agent, there is theoretically no reason why it wouldn’t be equipped to fulfill various professional tasks, too.

Increasingly, marketing and advertising professionals are assessing the ways in which AI agents might supercharge – or hamper – their own work.

The first and most straightforward advancement will be seen in operational efficiency at brands and agencies, according to Henry Cowling, chief innovation officer at S4 Capital agency Monks. He points to fintech company Klarna as a prime example. “They claim to have reduced marketing spend while driving growth by leveraging agents.”

It’s not just brands embracing the technology; agencies, too, are adopting agents to streamline operations and drive efficiencies. WPP’s chief technology officer and interim chair of AKQA Stephan Pretorious tells The Drum that the organization views agents “as digital coworkers who can increase the productivity of all our staff and help us get to better work faster.” He believes agents will prove to be “critical in terms of future agency competitiveness.”

The longer-term potential may be sexier, Cowling suggests, as AI agents could help advertisers reach target audiences at different parts of the customer journey with custom-tailored messaging and creative. “Most creative focus and resources [today] go into ‘big idea’ campaigns, leaving 80% of customer touchpoints underdeveloped,” he says. “With agentic workflows, brands can address this imbalance. We can deliver a ‘big idea’ for every audience, perhaps even every customer.”

Nonetheless, there are pitfalls ahead for advertisers.

The most glaring is the likelihood that an uptick in agent adoption among consumers will translate to reduced ad impressions for brands across the open web.

“There’s no interaction with the banner ads [on different websites] if you’re having your AI agent go and book your table on OpenTable,” explains Jeff MacDonald, who helms the AI Lab at Mekanism.

It’s a problem underscored by Arthur Fullerton, global chief technology officer at Havas CX. “Folks are looking to save time. Folks are looking for utilities that provide value … [and as such, AI agents are] removing the need to go to a destination website to carry out any action. The days of clamoring for people’s attention are getting harder.”

In the search landscape, this phenomenon has already taken root. The proliferation of AI chatbots and tools like Google’s AI Overviews is displacing search traffic for publishers, as AI programs’ direct response to user queries reduces the need to visit publisher websites for information. The shift is disrupting ad-driven revenue models tied to search traffic and is forcing publishers to reevaluate their content strategies.

Now, AI agents may dial up the pressure on publishers.

“It just becomes even more about SEO in an organic sense, where it’s like, ‘I hope I’m the best Walmart-versus-Target option when the OpenAI Operator goes and tries to find the best place to buy a pair of shoes,’” says Mekanism’s MacDonald.

In time, this shifting paradigm could fundamentally rearrange marketers’ and advertisers’ priorities.

“So much of the web is built around being human readable, and a lot of the advertising that we do is built around persuading humans,” MacDonald says. “Is there going to be a new world of advertising around persuading AI to make choices?”

Perhaps there’s even a mind-bending possible future state in which AI agents are programmed to market autonomously to other AI agents – cutting the human choice out altogether.

In any case, the upheaval will be significant, Havas CX’s Fullerton predicts. “Brands are going to have to modify what a destination website or what an online experience looks like,” he says, “and identify different opportunities to advertise.”

Luckily, all hope is not lost for advertisers, thanks in part to the fact that agents may be able to supply consumer data that will prove useful for ad targeting and measurement, Fullerton says. “Whether they’re standalone applications [or they] are built into SaaS products like Salesforce or AWS, these agents are still going to have a lot of information about the consumer, and that information is still going to be collected in some way and allow for different avenues [of marketing and advertising].”

OpenAI requires access to private information like email addresses and passwords when fulfilling some tasks with Operator, though the company has said that it does not retain this kind of data. OpenAI is, however, observing how Operator engages with users and the internet – information that it plans to use to adapt future versions of the program.

Despite the growing hype, technical hurdles may still impede the scaled deployment of AI agents for both consumer and enterprise use cases. Agents like Operator and Anthropic’s Claude Computer Use are still prone to reliability issues, including getting stuck in loops or generating inaccurate outputs. However, OpenAI’s decision to unveil Operator as a research preview first suggests a measured approach to managing expectations while collecting feedback to refine its capabilities.

Altman has previously indicated that artificial general intelligence – OpenAI’s ultimate target – could emerge when AI systems achieve a threshold of profitability and productivity akin to highly skilled human professionals. Whether Operator fulfills that vision or merely provides a glimpse into what’s possible, its launch signals a major shift in how AI tools might shape the future of work.