HR软件提供商Rippling推出了AI Spend Console工具，用于追踪和控制企业的AI支出。[1]这一工具的推出源于该公司在年初大规模投入AI后所暴露的成本问题。[1]
Rippling在内部遭遇了严重的AI超支。[1]该公司的R&D部门AI token支出曾高达员工薪酬预算的40%，月环比增长幅度达80%，其中一名工程师的月支出甚至达到50,000美元。[1]同时，公司内仅有10-15%的员工驱动了总AI支出的60%。[1]CFO Adam Swiecicki在3月的高管会议上首次提出这些数据。[1]
通过部署AI Spend Console和AI网关进行成本优化，Rippling取得了显著成效。[1]公司将token支出从40%降至约15%，而token使用量仍保持在高位——4月峰值使用605亿tokens，到7月仍使用600亿tokens，但同期成本反而下降37%。[1]此外，Rippling还发现采用成本更低的模型也能实现类似效果，例如Z.ai的GLM 5.2模型比前沿模型便宜85%且性能近乎相同。[1]
HR software provider Rippling has introduced the AI Spend Console, a tool designed to track and control enterprise artificial intelligence expenditures, following significant overspending within the company itself [1]. The initiative emerged after Rippling discovered severe cost overruns among its workforce in the early months of the year, with the R&D department's AI token spending reaching 40% of the employee payroll budget and experiencing month-over-month growth of 80% [1]. In one notable case, a single engineer's monthly AI expenses reached $50,000 [1].
The company identified a concentration problem in its spending patterns, finding that approximately 10–15% of employees were responsible for 60% of total AI expenditures [1]. CFO Adam Swiecicki presented these findings at an executive meeting in March [1]. Through implementation of the new tool combined with an AI gateway for cost optimization, Rippling successfully reduced its token spending ratio from 40% to approximately 15% [1]. While maintaining substantial usage levels—approximately 600 billion tokens in April and 600 billion tokens in July—the company achieved a 37% cost reduction between those months [1]. The improvement was partly enabled by switching to more cost-effective models, such as Z.ai's GLM 5.2, which offers 85% lower pricing than frontier models while delivering comparable performance [1].