Surging Use of ‘Pretty Crazy’ Tokens Challenges Executives’ AI Investments

At the software firm 8×8, team members are utilizing Anthropic’s Claude for tasks like drafting emails, evaluating customer feedback, and coding. However, their increasing dependence on the AI chatbot has not alarmed the finance department. In contrast to other Silicon Valley firms like Meta, Uber, and Salesforce, which have voiced concerns about the rising expenses associated with generative AI tools and implemented usage limits, 8×8 claims it remains profitable.

In the last 18 months, the company estimates it has saved around $5 million annually by discontinuing subscriptions to numerous software and educational resources that it found redundant, partly because Claude could offer similar functionalities. According to Joel Neeb, the company’s chief transformation and business operations officer, 8×8’s annual bill for Claude is significantly lower than that savings figure.

Neeb anticipates that savings and expenses will eventually balance out as 8×8 motivates more employees to embrace AI and integrates the technology into more complex tasks. For the moment, a considerable gap remains, which “makes my chief financial officer happy,” he shares with WIRED. He did not disclose the total expenditure on generative AI.

As companies collectively invest hundreds of millions of dollars into AI tools for coding, marketing, and customer support, a new fascination has surfaced in the tech industry: “tokenomics,” or how to handle the escalating costs of AI usage. (Tokens measure the volume of content an AI model analyzes and produces.)

Last month, the CEO of Royal Bank of Canada revealed that its token usage skyrocketed by 500 percent in the last six months. At Cisco, a third of its workforce is engaging with an internal AI chatbot daily, prompting CEO Chuck Robbins to note on an earnings call that “the token usage is getting pretty, pretty crazy.” Some leading engineers at analytics software company Amplitude are reportedly “spending thousands of dollars each month or more on tokens,” according to CEO Spenser Skates. Aaron Levine, Box’s CEO, mentioned that “the token budgeting conversation has absolutely taken over as one of the most important” and “heated” discussions.

Approximately 300 companies raised questions or concerns regarding AI tokens during their earnings calls or in public forums with financial analysts in April and May, as per a WIRED review of transcripts from the data provider AlphaStreet. Although this represents a small fraction of the thousands of calls made during that period, only 93 companies noted “token” in April and May of the previous year.

Executives from various companies stated they are either developing or seeking to acquire systems that can monitor token usage and select the most cost-effective model for specific prompts. Others indicated they are still trying to find a balance between hiring additional staff and increasing their budgets for tokens to meet their objectives.

Software has seldom been inexpensive, but the latest wave of AI tools is putting unusual pressure on C-suites for various reasons. Prices continue to fluctuate. Every month sees the launch of new models that are more powerful—and pricier—than their predecessors. Moreover, achieving organization-wide buy-in for new workflows has been challenging, resulting in productivity gains for one team causing bottlenecks for another.

20 Percent

Nonetheless, some companies still encourage their employees to utilize AI without concern over expenses. In April, Baseball Lifestyle 101, a clothing brand based in Long Island, New York, which anticipates generating $250 million in sales this year, instructed around 50 of its top managers to invest roughly 20 percent of their salary in AI tokens each month.

Co-founder and chief strategy officer Bill Rom shared with WIRED that costs are expected to surpass $100,000 a month by year-end, but the investment is already yielding returns. Claude recently assisted in securing a $1 million order by revealing that a retailer was running low on certain sizes of the company’s popular ice-cream-patterned shorts. “That’s a day and a half of work that can now be completed in one or two hours, potentially generating me eight figures of additional revenue over 12 months,” Rom explains.