Ai agents are too cheap for our good

Zhang says that his company makes money for every single conversation after excluding some general costs, but has refused to comment on the overall profitability of the startup. With $ 100 million collected From the venture capitalists including Andreessen Horowitz and Accel, Decagon has the flexibility to give priority to growth compared to profitability. “If we could make a price more, it’s always like a ‘what if?’,” He says. “But in general we are quite happy at the moment.”

“So cheap”

Erica Brescia, CEO of the investment company Redpoint Ventures, had an epiphany regarding the prices of artificial intelligence agents last month. The price of $ 250 on the new Ultra Piano di Google has stunned it. “All this is so cheap,” he remembers having thought. “It is disproportionate for the value that people are getting.” He felt that a price of at least double would make more sense. (That same week, said the CEO of Nvidia Jensen Huang Stratechery who would have assumed an artificial intelligence agent for $ 100,000 per year “in a heartbeat”. )

Previously, Brescia worked as Chief Operating Officer of Github, who helped to establish the level for artificial intelligence prices. The coding assistant of the Github Copyloot started $ 10 per month in 2022months before the chatgpt debut. Brescia says that Github went with a price that would have attracted a critical mass of users. The goal was to collect data to improve the service and the parent company of Github, Microsoft, did not matter to miss a new tool to make it happen. In reality, a 100 -time price would better reflect the value that Copilot provides software developers, Brescia estimates. (Kyle Daigle, Github’s Chief Operating Officer, tells Wired that the company’s goal is to support, not replace, developers and that “prices reflect a commitment to democratize access to powerful tools”.

Today, Copilot is at the top of $ 21 per month. And similar tools followed his advantage, including Zed, who received $ 12.5 million in funding from Redpoints and others. In May, the company began loading a minimum of $ 20 per month for a code of code assisted by the AI ​​who built from scratch.

The CEO of Zed Nathan Sobo provides that artificial intelligence companies are charged more over time because the current price models are not sustainable. But relating to humans, he wants to maintain artificial intelligence agents at affordable prices so that anyone can use them to increase their work, develop better software and create new jobs. “I want more intelligence at my disposal at a low cost as possible,” he says. “But for me, included in this is potentially a junior engineer who uses this technology, ideally at a low cost as possible.”

The Zhang of Decagon feels in the same way for the AI ​​coding tools. “Would we pay more? Marginally? Yes,” he says. But “$ 2,000? Probably not.” He adds “hunger for good engineers is infinite”.

AI entrepreneurs suggest that agents could command higher prices if they were easier to set up and more reliable to use. For example, Nancedita Giri, a senior software engineer who worked in Amazon, Meta and Microsoft, says he would pay thousands of dollars a year for a personal artificial intelligence assistant. “But rigorous conditions apply: you can’t be frustrated using it,” he says.

Unfortunately, that day it seems far away. As a personal project, Giri tried to develop an artificial intelligence agent capable of preventing psychological burnout. “He just canceled all my meetings,” he says. Certainly a solution, but not the ideal one.

Now, some companies are hiring “to architects” to help supervise agents systems and reduce gaffes. The question is those who will take care of those roles in the future if today’s workers are cut by opportunities. Simon Johnson, an economist of the Massachusetts Institute of Technology, does not expect companies to take into account the social cost of the interruption of the career in making their decisions on prices. It suggests to governments that reduce taxes on wages for entry-level roles to encourage intake. “The right lever to pull is the one that reduces the costs to employers,” says Johnson.

Arrigoni is choosing a third path. In Loti Ai, he gave priority to the constant hiring of junior engineers and did not use coding tools. If the apocalypse of work arrives, “I don’t want to be guilty,” he says.