AI token costs strain tech budgets as companies chase efficiency gains

At a glance:

• 8x8 saved $5 million annually by replacing software tools with Anthropic's Claude chatbot
• Royal Bank of Canada saw token usage surge 500% in six months amid enterprise AI adoption
• Baseball Lifestyle 101 allocates 20% of top managers' salaries to AI tokens monthly

The cost of going all-in on AI

At 8x8, a communications platform developer, employees have embraced Anthropic’s Claude for tasks ranging from email drafting to code generation and customer feedback analysis. Unlike peers such as Meta, Uber, and Salesforce—who have publicly flagged rising AI expenses—8x8 reports net savings. Over 18 months, the company eliminated dozens of software and educational tool subscriptions, estimating $5 million in annual cost reductions. Its current Claude spending remains "well below" that figure, according to Joel Neeb, 8x8’s chief transformation and business operations officer.

However, Neeb anticipates costs will rise as adoption expands. The recent release of Claude Opus 4.8, which costs 1.7 times more than Anthropic’s February model, has sparked internal discussions about potential usage caps. While no decisions are final, future access to premium models may require demonstrating older versions cannot meet objectives. "Can we downgrade the model a little bit and still get the same outcome?" Neeb asks. Despite this, 8x8 shows no signs of retreating: customer satisfaction metrics have improved, and revenue has grown for four straight quarters, with AI-driven sales acceleration playing a suspected role.

Industry-wide token anxiety

The term "tokenomics"—managing AI usage costs—has become a boardroom obsession. Tokens quantify the content processed by AI models, and their consumption is escalating rapidly. Royal Bank of Canada’s CEO revealed a 500% token usage spike over six months. Cisco’s Chuck Robbins noted that a third of employees use internal AI tools daily, calling token consumption "pretty, pretty crazy." At analytics firm Amplitude, some engineers spend thousands monthly on tokens, while Box’s Aaron Levine calls token budgeting a "heated" priority.

A WIRED review of AlphaStreet transcripts found roughly 300 companies raised AI token concerns in April-May earnings calls—up from 93 a year prior. Executives are racing to implement monitoring systems and optimize model selection. Yet balancing AI investment against hiring remains unresolved. "Software has rarely come cheap," the article notes, but generative AI introduces unique pressures: volatile pricing, frequent model upgrades, and uneven organizational adoption creating workflow bottlenecks.

Baseball Lifestyle 101 bets big on AI

New York-based Baseball Lifestyle 101, projecting $250 million in 2024 sales, took an unconventional approach. In April, it instructed 50 senior managers to dedicate 20% of their salary-equivalent to AI tokens monthly. Bill Rom, the company’s cofounder and chief strategy officer, expects costs to exceed $100,000 monthly by year-end—but sees immediate returns. Claude identified a retailer’s inventory shortage of ice-cream-patterned shorts, securing a $1 million order. "That’s a day and a half of work that can now happen in an hour or two," Rom says, potentially driving eight-figure revenue gains annually.

The chatbot also streamlines financial reporting and photoshoot planning, reducing junior staff needs. Rom emphasizes inspiring AI use before enforcing financial controls. Meanwhile, 8x8’s 1,800 employees regularly check a dashboard tracking team-wide Claude usage. Neeb describes it as non-punitive: "It's really just so that we all stay tightly packed in this journey."

Challenges in enterprise AI adoption

Two years ago, 8x8 trained all employees on ChatGPT and Google Gemini before standardizing on Claude. Management warns that AI fluency is now mandatory: "If you're not using AI in some capacity for your role, then you're missing the opportunity to go faster," Neeb says. Other firms report mixed results—Amazon and Meta workers allegedly misuse AI tools out of obligation or complacency, prompting criticism of inefficiency.

Neeb targets underperforming teams like sales and finance (28% of staff but only 15% of token use) through initiatives like AI hackathons. He personally uses Claude to automate a daily email summarizing AI tips, recently optimizing the process to cut token consumption by 80% after identifying inefficiencies. As companies navigate AI’s double-edged promise of productivity and expense, the focus shifts from adoption to sustainable integration.

What's next for AI cost management

The surge in token usage reflects broader enterprise AI maturation. Companies like 8x8 and Baseball Lifestyle 101 exemplify early success stories, but scaling these gains requires tighter cost controls. Expect more firms to adopt hybrid strategies: leveraging cheaper models where possible while reserving premium tools for complex tasks. Monitoring systems and usage quotas may become standard as CFOs demand accountability. For now, the AI arms race continues, with efficiency gains offsetting costs—but not indefinitely.