Rippling's CFO Shock: How Runaway Token Bills Spawned a New AI Cost-Control Product

Rippling unveiled AI Spend Console this week, a product designed to help companies track, analyze, and contain their AI usage costs at the individual employee and team level. The launch comes directly out of a crisis the HR software company experienced earlier this year, when its own internal AI spending spiraled far beyond what leadership had anticipated. The tool links AI token expenditure to measurable work output — things like pull requests and lines of code — so managers can judge whether high spenders are actually delivering results or just generating expensive AI slop.

The origin story is striking. In March, Rippling CFO Adam Swiecicki presented figures to the executive team showing the company was on track to spend the equivalent of 40% of its R&D payroll on AI tokens alone. At the time, token spending was growing at roughly 80% month-over-month, which meant the company was hurtling toward a scenario where token costs could approach 90% of its entire engineering compensation budget within a year. Chief Product Officer Matt MacInnis described the room's reaction simply: 'We were incredulous.'

A subsequent internal audit revealed that 10 to 15% of employees were responsible for around 60% of total AI spend, with one engineer alone running up a $50,000 monthly tab. Rippling moved quickly, negotiating spending caps with tool providers including Cursor, OpenAI, and Anthropic. Executives also identified a structural problem: employees were defaulting to the newest and most expensive frontier models regardless of task complexity, and the inference providers had little motivation to help fix that. MacInnis said those vendors 'have every incentive for it to be a runaway expense.'

Rippling's solution included building its own AI gateway that routes prompts to the most cost-effective model suited for each task. MacInnis noted that cheaper models often perform comparably to frontier options — Rippling CEO Parker Conrad had recently flagged that Z.ai's GLM 5.2 delivered nearly identical coding performance to top-tier models at 85% lower cost. SpaceX's Grok also ranked highly in Rippling's internal benchmarks. The gateway is central to the AI Spend Console; companies using a third-party gateway can still access dashboards and reporting features, but the cost-governance controls require Rippling's own routing layer.

The results have been measurable. Rippling reduced its token spend from roughly 40% of headcount budget down to about 15%, without actually curtailing AI usage volume. In July, internal token consumption hit approximately 600 billion — near the March peak — yet July's total cost was only 37% of what April's bill had been. The company also identified high-performing AI users internally and designated them 'AI captains' to spread effective practices across teams. MacInnis acknowledged that extending smart AI usage beyond engineering remains a work in progress.