Rippling Launches AI Spend Console After Token Costs Hit 40% of R&D Pay
Rippling cut AI token costs by routing work across models, then launched AI Spend Console to tie employee usage and spending to measurable productivity.
Summary
In the week of August 7, 2026, Rippling unveiled AI Spend Console to curb tokenmaxxing after CFO Adam Swiecicki warned executives in March that AI tokens were tracking at 40% of R&D compensation, costing millions. Spending was rising 80% month over month and, if sustained, would approach 90% of R&D pay the next year. An audit found 10% to 15% of employees generated about 60% of AI spend, including one engineer at $50,000 monthly. Chief Product Officer Matt MacInnis said Rippling capped spending with Cursor, OpenAI and Anthropic after workers defaulted to the newest, costliest frontier models.
The console maps costs and productivity by employee, team and role, flags expensive engineering work repeatedly redone in code review, and scores prompts, code lines, pull requests, output and spend. Its gateway routes tasks among models; other gateways can feed the console, but Rippling’s is required for spending controls. CEO Parker Conrad said internal benchmarks ranked SpaceX’s Grok best overall, while Z.ai’s Chinese coding model GLM 5.2 delivered nearly identical performance at 85% lower cost; Databricks also promotes it, and SpaceX owned Cursor offers Grok and dozens of models. Routing cut token spend from 40% to about 15% of headcount budget. July usage reached 600 billion tokens, near the 605 billion warning month peak, but cost 37% of April’s total, partly by avoiding models such as Fable for grammar edits. Rippling appointed effective users as AI captains and is testing onboarding automation for mailing data and reconciliation, measuring customers onboarded. MacInnis said broader access depends on tying G&A and customer facing token use to productivity. The console is included for Rippling HR subscribers with added usage fees, or sold standalone for integration with another HR system.
Positives
- Routing reduced Rippling’s AI token spending from 40% to about 15% of its R&D headcount budget without curtailing usage.
- July’s 600 billion tokens cost only 37% of April’s total, despite approaching the 605 billion warning month peak.
- GLM 5.2 delivered nearly identical internal performance to frontier models at 85% lower cost, according to Parker Conrad.
- AI captains now help colleagues adopt effective practices, while onboarding teams are testing mailing data and reconciliation automation.
- Rippling HR subscribers receive AI Spend Console with usage fees, while other companies can buy it standalone.
Risks & concerns
- AI tokens were consuming millions and tracking at 40% of R&D compensation when Adam Swiecicki raised the alarm in March.
- Spending was growing 80% monthly and could have reached 90% of R&D compensation the following year.
- Just 10% to 15% of employees generated about 60% of AI spending, with one engineer consuming $50,000 monthly.
- Employees routinely selected the newest and most expensive frontier models regardless of task requirements.
- Employees outside engineering may lose broad AI access unless Rippling can connect token consumption to measurable productivity.