Companies track AI token costs as workplace chatbot use grows

In short: As more employees use AI chatbots at work, companies are paying closer attention to “tokens,” the unit that often determines AI costs.

What's going on

More businesses are running into a new budgeting problem tied to AI chatbots like Anthropic’s Claude. Many AI services charge based on “tokens,” which are small pieces of text the system reads and writes (like paying by the word, instead of a flat monthly fee). As usage rises, costs can rise quickly.

WIRED reports that around 300 companies brought up AI token concerns in earnings calls or public talks with analysts in April or May. That is up from 93 companies mentioning “token” in the same period a year earlier. Leaders at companies like Royal Bank of Canada, Cisco, Amplitude, and Box have described token usage jumping fast, and some workers are spending thousands of dollars a month.

8x8, a communications software company with about 1,800 full-time employees, says it is still coming out ahead. The company estimates it saved about $5 million a year by canceling dozens of other software and training subscriptions, in part because Claude could handle similar tasks. Its annual Claude bill is “well below” that figure, according to Joel Neeb, 8x8’s chief transformation and business operations officer.

Still, 8x8 is watching costs as it rolls out a newer Claude model that is about 1.7 times more expensive than an earlier one. Neeb says the company may eventually set limits, or ask employees to prove they need the pricier model.

What to watch

Expect more “token budgeting” tools and rules inside companies, especially as newer, more expensive AI models arrive. For workers, this could look like tighter tracking, usage caps, and pressure to use cheaper options when they are “good enough.”

Source: Wired