E-commerce web sites may also attempt to pivot to the agentic consideration economic system, however the path ahead there may be unclear. Final yr, a group led by Akshit Kumar, a professor of operations administration on the College of Toronto Scarborough, investigated whether or not and the way brokers are influenced by the ways in which purchasing web sites current search outcomes. They discovered that brokers are much less prone to suggest gadgets that include a “sponsored” tag, which means that promoting favorable positions on outcomes pages may not be an effective way for websites to earn money if brokers turn into the first means that folks store on-line. Fashions did appear to have a desire for gadgets that appeared specifically positions in search outcomes—however whether or not they prioritized gadgets on the prime, backside, or center of the outcomes web page differed from mannequin to mannequin and between generations of a single mannequin. It may very well be difficult for e-commerce platforms to maintain up.
David Rothschild, an economist at Microsoft Analysis, says he’s hopeful that the transition to the agentic internet may put extra energy within the arms of the patron. Paying for on-line content material in the present day usually requires a dedication: We subscribe to {a magazine} or be a part of a YouTuber’s Patreon, and in flip we acquire entry to their full output. One various is micropayments that permit customers purchase content material à la carte, however the concept has foundered up to now as a result of paying article by article or video by video requires an excessive amount of administrative overhead. Brokers, although, may handle that overhead for us, letting us pay immediately for every bit of content material we devour and placing only a bit of money immediately within the arms of every particular person whose work we take pleasure in.
Different consultants share the same imaginative and prescient. Mustafaraj, the Wellesley pc scientist, anticipates that brokers may assist us escape from the algorithms that management a lot of the media we devour. “My hope is that we are able to have AIs that work for us, within the sense that they’ll change the feed the best way I would like—not the best way TikTok desires,” she says. And Varma, of Mozilla, imagines an web with out intermediaries like Airbnb and Doordash, which now extract tolls for facilitating interactions between shoppers and sellers. As a substitute, he imagines, maybe we are going to all have brokers that may securely and reliably interface immediately with, say, eating places’ brokers to get dinner delivered to us, freed from pesky service charges.
It’s an interesting concept, however the historical past of the net has been a narrative of consolidation and focus of energy and management, and that is perhaps the default path right here, too. When the private assistant agent OpenClaw exploded onto the scene in early 2026, it appeared, for a short second, that brokers is perhaps inbuilt a grassroots, distributed means. However OpenAI shortly snapped up OpenClaw’s creator, Peter Steinberger, and different frontier AI corporations set to work constructing their very own private assistant brokers.
Once we use these brokers, we’re on the mercy of how the businesses need their instruments to behave. “You might have this know-how that may simply eradicate all this communication friction,” Rothschild, the Microsoft economist, says. “You’re going to see corporations construct obstacles to be able to preserve the patron base that they’ve.”
Right here’s what that might appear to be: Let’s say Muse turns into the default means that many individuals discover data, store, and handle their digital lives. In precept, customers may deploy Muse to scour the web for the content material most useful to them, join immediately with distributors, and in any other case escape the walled gardens which have outlined the web for the previous couple of many years. However Meta may not need to allow them to try this. As a substitute, it’d design Muse to maintain customers inside its personal walled backyard—by, say, preferentially exhibiting customers content material from Meta social networks and recommending companies which have paid a price to be featured prominently in Muse outcomes. Meta and OpenAI didn’t reply to requests for remark for this story.
There are some advantages to this extra closed-off model of the agentic internet. Brokers will be highly effective instruments for hackers and scammers. The main corporations have safeguards in place to stop, or a minimum of mitigate, this misuse—for instance, they’ve instruments that may routinely establish when somebody is making a hacking-related request and block a mannequin from responding. In a completely open ecosystem, nonetheless, somebody would absolutely construct and disseminate a system with no such restrictions. Corporations even have a robust reputational incentive to stop their private assistant brokers from making main errors, resembling authorizing a big buy that the person didn’t need—and they also’ve put loads of effort into constructing dependable guardrails.
