AI agents are still too expensive for businesses and do not always operate at full capacity. According to two multimillionaires from the tech sector, with current expenses, replacing people with them will be difficult.
Investor Jason Calacanis said on the All-In podcast that he pays about $300 a day for the AI agent Claude from Anthropic, which helps manage his companies. That adds up to about $110,000 per year. At the same time, he says the bot is only used at 10-20% of its full capacity.
“At what point do tokens become more expensive than an employee’s salary?” Calacanis wondered, referring to model usage limits.
To use most AI services, users buy tokens, and with heavy usage their cost rises quickly.
Head of Social Capital Chamath Palihapitiya noted that he faces the same problem. According to him, at current prices, the models need to be at least twice as productive as a regular employee to justify the costs. He also suggested that companies may have to introduce a separate budget for AI usage.
See also: AI Agents Enter the Market. Stablecoins Are Preparing Infrastructure for Them
Mark Cuban found this argument the most convincing counterargument against the idea that AI will massively displace people from jobs. On Thursday, he wrote that the high cost of implementation was the most reasonable objection he had seen.
By his estimate, including tokens and technical maintenance, eight Claude AI agents Claude could cost $1,200 a day to perform the same volume of tasks as one employee.
Cuban doubted that such bots are really more than twice as productive as a person. He also emphasized that decision-making involves factors that are difficult to measure, such as team morale and ethical considerations.
Concerns about replacing people with AI have intensified in recent years. Some companies have already made layoffs, explaining that some tasks can be automated.
A study by Microsoft published in July showed that the professions most at risk are those related to intellectual work, as well as customer service and sales.
However, not everyone considers the threat of mass layoffs to be real. White House adviser on AI and cryptocurrencies David Sacks said in August that the concerns are exaggerated. According to him, models still require correct prompts and result verification to deliver real business value.
At the same time, consulting company McKinsey & Co emphasizes that the goal of such AI agents is full process automation. They should perform tasks from start to finish without constant human involvement.
For now, the main question remains: can AI agents become not only technologically advanced but also economically justified for business.
Stablecoins Could Become the Native Currency for AI Agents
The popularity of AI agents in the crypto community is growing. More and more, they are seen not as an auxiliary tool, but as market participants who can conduct transactions themselves.
Head of Circle Jeremy Allaire said last month that in five years, billions of AI agents will settle payments with stablecoins on behalf of users. According to him, these assets are best suited for automatic payments.
Co-founder of Binance Changpeng Zhao said back in January that cryptocurrency will become the native currency for AI agents. He believes that blockchain is better suited for their work than other technologies.
AI agents are already operating in several networks. On the Base network, which is a layer-two solution on Ethereum, the AIXBT project uses the Virtuals protocol to conduct micropayments and help close deals.
In the Fetch.ai ecosystem, the ASI Alliance manages assets and performs economic operations on behalf of users.
Model developers are also paying more attention to this area. On Wednesday, OpenAI introduced a new benchmark. It tests how well different models find vulnerabilities in smart contracts, fix them, and can use them in attack scenarios.
See also: Japan Bets on Debt and Why It Matters for Bitcoin Traders
The company noted that it is important to test such systems in an environment where real money is at stake.
“Smart contracts protect assets worth billions of dollars, and AI agents can change both attack tools and defense methods,” OpenAI said.
The intersection of AI and blockchain is gradually moving beyond experiments. More and more projects are testing how such agents can work with real assets and payments.
