AI
Rippling built an AI cost-control platform after burning millions on tokens in months
The HR software company discovered its R&D team was on track to spend as much on AI models as it paid employees, prompting an urgent overhaul in how it allocates models and monitors spending.
TechCrunch — AI
HR software vendor Rippling this week unveiled its new AI Spend Console, a tool designed to help organizations monitor and control their spending on AI services. The platform identifies which employees and teams are consuming the most AI resources and whether those costs are actually translating into meaningful productivity gains.
The shock that sparked action#
Rippling's initiative came after the company went all-in on AI adoption earlier this year, only to discover unsustainable spending patterns. In March, the finance team presented numbers that alarmed leadership: the company was on pace to spend 40% of its entire R&D headcount budget on AI tokens alone—effectively matching the compensation costs for that unit. With spending increasing 80% month-over-month, the trajectory suggested AI token costs could reach 90% of R&D payroll by the following year.
An internal audit revealed striking disparities: roughly 10–15% of employees were responsible for 60% of total AI spending, with one engineer consuming $50,000 monthly. "We were incredulous," Chief Product Officer Matt MacInnis told TechCrunch.
Rethinking the model portfolio#
Rather than cutting AI access, Rippling focused on using smarter procurement. The company set spending caps with its AI providers—Cursor, OpenAI, and Anthropic—but discovered a more fundamental problem: employees defaulted to the latest, most expensive frontier models for every task.
MacInnis noted that inference providers have little incentive to help customers control costs. To address this, Rippling developed an internal AI gateway that routes prompts to the most cost-effective model for each task. The company found that cheaper alternatives often matched the performance of premium options. Rippling's founder noted that Chinese model GLM 5.2, priced at 85% less than frontier models, delivered nearly identical results.
Tracking ROI and building accountability#
The new platform provides dashboards measuring employee AI activity against output—tracking metrics like lines of code, pull requests, and prompts per day relative to spending. After implementing the system, Rippling cut its token spending from 40% down to 15% of R&D payroll while maintaining total token usage at similar levels through better model selection.
The company also created "AI captains"—employees identified as effective AI users who mentor their peers in best practices.
