Rippling builds AI spend tracker after burning millions on tokens

Rippling, an HR software provider, this week unveiled AI Spend Console, a product designed to track and constrain AI spending while measuring whether increased token use actually translates to employee productivity. The tool's central feature maps spending by individual employees, teams, and roles, then evaluates whether high-spending users are genuinely more productive or simply generating low-quality output—what the company calls "AI slop." Rippling even highlights that the dashboard will flag "which engineers have high AI spend whose peers frequently ask them to redo work in code reviews."

The product's origin story began in March 2026, when Rippling's leadership discovered the company was on a collision course with runaway AI costs. Chief Product Officer Matt MacInnis recalls the shock when CFO Adam Swiecicki presented the R&D organization's token spending trajectory: Rippling was on track to burn 40% of its entire R&D headcount budget on AI inference tokens—meaning it would spend as much on tokens as on the salaries of 40% of its engineering workforce. Spending was climbing 80% month-over-month, and if unchecked, would reach 90% of the headcount budget within a year. The company's peak monthly token consumption hit 605 billion tokens in April. "We were incredulous," MacInnis told TechCrunch. Management launched an urgent investigation into the spending and what productivity it was generating.

Analysis revealed two shocks. First, roughly 10–15% of employees were responsible for approximately 60% of total spend. One engineer alone was spending $50,000 a month. Second, the underlying problem was structural: employees routinely selected the most recent and expensive frontier models for all tasks, from routine code formatting to complex reasoning. Rippling negotiated spending caps with its AI providers—Cursor, OpenAI, and Anthropic—but discovered that inference providers have little incentive to help customers control costs. As MacInnis noted, "The inference providers, like Anthropic and OpenAI, have absolutely no incentives to help you control your spend. They have every incentive for it to be a runaway expense."

Rippling's solution involved two moves. First, it built its own AI gateway that routes prompts to the most cost-effective model for each task rather than always defaulting to frontier options. Rippling founder Parker Conrad noted in internal benchmarks that while SpaceX's Grok emerged as the strongest all-around performer, Z.ai's GLM 5.2—a Chinese open-weight model—delivered nearly identical performance while costing 85% less. GLM 5.2 has become a favorite among tech companies for coding tasks. Second, Rippling created AI Spend Console dashboards that score attributes including prompts per day, work output (lines of code and pull requests), and token spend—tools MacInnis jokes prevent "the sales team from doing grammar updates using Fable" (a reference to expensive frontier inference).

The results were striking. Rippling reduced token spend from 40% of headcount budget down to about 15%. More tellingly, while July 2026 usage reached 600 billion tokens—matching April's peak—the cost of July's token spend was only 37% of April's cost, purely because the company was routing to more effective, cheaper models. The company also appointed "AI captains"—employees who use AI effectively—to guide the rest of the organization.

However, Rippling notes that technology alone is insufficient. The company faces a remaining challenge: extending AI governance beyond engineering. Software engineers have been the primary users so far, and Rippling is working to measure productivity gains in customer-facing functions like onboarding automation and data reconciliation. MacInnis emphasized: "We have to be able to link token consumption in G&A functions and in customer-facing functions back to productivity. If we can't do that, all bets are off on any of this stuff being available to the broader employee base." This suggests that if enterprises cannot quantify AI's productivity impact, employee AI access may not become as universal as email or Slack.

AI Spend Console is included for Rippling's existing HR subscribers (with additional usage-based fees) and can also be purchased standalone and integrated with third-party HR systems, though enterprises that want Rippling's spending-governance features must use its gateway.