In 2007, Luke Arrigoni an AI entrepreneur, made $63,000 in his first job as junior software developer. He says that AI tools today cost only $120 per year and can write better code than his did then.
He doesn’t like the numbers. Arrigoni runs Loti AI – a company which helps Hollywood stars to find deepfakes. He is concerned that low-priced AI software encourages companies to eliminate entry level roles. He wants to change the incentive structure, so that people’s careers do not end before they even begin. “If you make AI systems more costly, you have an incentive to hire someone who is just starting out,” says he.
The fear that AI will transform–or eliminate–jobs is a perennial concern. The concern is growing as the demand for AI agents increases. These AI systems are now able to make sales calls and code software, tasks that were once reserved for human beings.
The situation isn’t bad yet. Hiring platform ZipRecruiter According to estimates summer internships in America have returned to a level similar to what they were prior the pandemic. But this could change in the near term. OpenAI CEO Sam Altman, who spoke at the Snowflake summit in San Franciscolast week, compared AI tools today to interns. He said that the next-generation technology will be like a “more experienced” worker. Altman said that in some companies, managers are already overseeing “a bunch” of agents the same way they used to supervise “relatively younger employees”.
OpenAI talked about mitigation measures like reskilling to stave off an potential jobs crisis, but it hasn’t discussed charging higher prices for their services to slow down the transition to AI. Arrigoni is on edge. Even after adding in the most expensive features, AI coding agent cost a fraction as much as a junior engineer. Arrigoni says that if inexperienced workers are unable to get a job, they may not gain the experience needed to lead future teams, whether human or machine.
OpenAI has not responded to a comment request.
AI pricing Since ChatGPT launched in 2022 as a free bot, it has fluctuated and triggered a boom in AI. In general, AI companies offer free tiers with limited usage, and the prices of basic tiers are declining. The top-tier plans with the latest features are more expensive, but not so much that they discourage adoption or generate profits for the companies providing them.
Startup executives attribute low prices among AI providers to intense competition. Ajit Ghuman is the CEO of Monetizely, a pricing strategy company. He says that “their only chance to win is through mass adoption.” This means AI companies must charge the same affordable price as their competitors. Ghuman says that prices are unlikely to rise significantly unless there is a major problem with electricity or GPU shortages, or if one company dominates the AI market.
Decagon is a San Francisco-based startup that sells customer service chatbots used by retailers and technology companies. It charges $1 or less Per conversation– roughly half the cost of human assistance. Decagon believes that even though the chatbot is more effective in some cases than a human, its clients will never pay more. Jesse Zhang, CEO of Decagon, says that the main reason to invest in AI technology is efficiency. “You’re going be less than human work.” “That’s like the point of technology.”
Zhang
says that his company makes money from each conversation, after subtracting certain overhead costs. He declined to comment on overall profitability. Decagon can prioritize growth over profitability with $100 million raised from venture capitalists Andreessen Horowitz, Accel and Andreessen. “It’s always a question of ‘what if’ as to whether we could be charging more,” he says. “But in general, we’re pretty satisfied right now.”
‘So Cheap’
Erica Brescia had an epiphany last month about AI agent pricing. She was shocked to see the $250 price tag for Google’s new AI Ultra Plan. She remembers thinking, “All of this is so cheap.” “It is disproportionate to the value that people are receiving.” She felt that a price at least twice as high would be more reasonable. Brescia was previously the chief operating office of GitHub. This helped set the standard for AI pricing. GitHub’s Copilot coding assistance started at a href=””https://github.blog/news-insights/product-news/github-copilot-is-generally-available-to-all-developers/” “rel=””nofollow noopener” “target=””_blank””>$10 a monthly in 2022, months before ChatGPT debuted. Brescia said GitHub chose a price to attract a critical mass. Microsoft, GitHub’s parent, was willing to take a loss to gather data for the service. Brescia estimates that a price 100-fold higher would better reflect the value Copilot offers to software developers. (GitHub’s chief operating officer Kyle Daigle told WIRED that GitHub’s goal was to support developers, not replace them, and that “pricing reflects a dedication to democratizing the access to powerful tools.”
) Today, Copilot is priced at $21 a monthly. Similar tools have followed suit, including Zed which has received funding of $12.5 million from Redpoint, among others. In May, the company began charging $20 per month for an AI-assisted editor that it had built from scratch.
Zed’s CEO Nathan Sobo believes that AI companies will charge more in the future because current pricing models won’t be sustainable. He wants AI agents to be affordable for humans so that anyone can use them as a tool to enhance their work, create better software, or create new jobs. “I want to have as much intelligence as I can at a low cost,” he says. “But I think that includes a junior engineer who could use this technology at the lowest possible cost.”
Decagon’s Zhang feels the exact same way about AI-coding tools. “Would you pay more? Marginally? “Yes,” he replies. But “$2,000?” “Probably not.” He added, “the hunger for engineers is endless.”
AI entrepreneur suggest that agents could command a higher price if they were easier and more reliable to set up. Nandita Giridi, a senior engineer who has worked for Amazon, Meta and Microsoft, said she would pay thousands of dollar per year for an AI personal assistance. “But strict rules apply – you can’t become frustrated using it,” she says.
Unfortunately that day seems far away. Giri developed an AI agent to prevent burnout as a personal project. “It canceled all of my meetings,” she explains. It’s a good solution, but it’s not ideal.
Some companies are now hiring “AI architects” who will help oversee agentic system and reduce gaffes. The question is, who will fill these roles in the future, if young workers are denied opportunities today? Simon Johnson, an economics professor at the Massachusetts Institute of Technology (MIT), doesn’t think companies will consider the social cost of career disruption when making pricing decisions. He suggests that governments lower payroll tax for entry-level jobs to encourage hiring. Johnson says that the best lever to pull would be one that lowers costs for employers.
Arrigoni chooses a third way. He has hired junior engineers at Loti AI and has not used AI coding tools. If the job apocalypse happens, “I do not want to be at blame,” he says.
