Chang says that after excluding certain overhead costs, his company makes money in individual conversations, but he refused to comment on the overall profitability of the startup. With $100 million raised from venture capitalists including Andreessen Horowitz and Accel, Decagon has the flexibility to prioritize growth over profitability. “Whether we can price more or not, it's always like, 'What if?',” he says. “But in general we are pretty happy now.”
“Very cheap”
Erica Brescia, managing director of investment firm Redpoint Ventures, had a terrible thing about pricing for AI agents last month. Google's new AI Ultra Plan's $250 price tag surprised her. “All of these are very cheap,” she recalls the idea. “It's disproportionate to the value people are getting,” she felt that at least twice as much price made more sense. (That same week, Nvidia CEO Jensen Huang told Stratechery he would hire an AI agent for $100,000 a year “in a heartbeat.”
Previously, Brescia worked as Chief Operating Officer at Github, which helped set up the bar for AI pricing. Github's Copilot Coding Assistant began in 2022 for $10 a month, a few months before ChatGpt's debut. Brescia says Github has brought prices to attract key masses of users. The goal is to collect data to improve our services, and Microsoft, the parent company of Github, didn't mind losing money with new tools to make it happen. In reality, 100 times higher prices have now better reflected the value that Copilot offers to software developers, Brescia estimates. (Kyle Daigle, Chief Operating Officer at Github, told Wired that the company's goal is to support developers rather than replace them, and “pricing reflects its commitment to democratizing access to powerful tools.”
Today, Copilot is tops at $21 a month. The lead was also followed by similar tools, including Zed, who received $12.5 million in funding from RedPoint and others. In May, the company began charging more than $20 a month for an AI assist code editor built from scratch.
Zed CEO Nathan Sobo hopes AI companies will charge more bills over time, as the current pricing model is not sustainable. However, compared to humans, he wants to keep AI agents affordable, allowing anyone to use them to increase jobs, develop better software, and create new jobs. “I want to use my intelligence at my own cost as much as possible, at the lowest possible cost,” he says. “But I could potentially include junior engineers who use this technology. Ideally, it's as low as possible.”
Decagon's Zhang feels the same way about AI coding tools. “Are we going to pay more? Slightly? Yeah,” he says. But “$2,000? Probably not.” He adds, “The hunger for a good engineer is endless.”
AI entrepreneurs suggest that agents can direct higher prices if they are easier to set up and more reliable to use. Nandita Giri, a senior software engineer who worked for Amazon, Meta and Microsoft, says he pays AI personal assistant thousands of dollars a year. “But strict conditions apply. You can't get frustrated by using it,” she says.
Unfortunately, that day feels far away. As a personal project, Giri has attempted to develop an AI agent that can prevent psychological burnout. “It just cancelled all my meetings,” she says. It's certainly a solution, but not ideal.
Now, some companies are hiring “AI Architects” to help oversee the agent system and reduce rudeness. The question is, if early career workers are separated from today's opportunities, who will take on those roles in the future? Simon Johnson, an economist at the Massachusetts Institute of Technology, does not expect companies to take into account the social costs of career disruptions in determining pricing. He proposes that the government cut payroll taxes on entry-level roles to encourage employment. “A good lever to pull reduces the cost of an employer,” says Johnson.
Arrigoni has chosen the third pass. At Loti AI, he steadily prioritizes hiring junior engineers and does not employ AI coding tools. When the work apocalypse comes, he says, “I don't want to be negligent.”
