.png){kind=logo}
Rippling has launched AI Spend Console after its internal AI token costs climbed into millions of dollars. The product tracks spending by employee, team, and role. It also compares usage with work measures, including code reviews and pull requests.
The HR software company developed the system after reviewing rapid spending growth inside its research and development unit. Rippling says the console helps companies control costs while keeping AI tools available for productive work.
In March, Rippling’s finance team found that AI tokens could consume 40% of its R&D headcount budget. Monthly spending had grown by 80%, putting the projected annual cost near 90% of that payroll budget. Chief Product Officer Matt MacInnis described executives as “incredulous” when they reviewed the figures.
Rippling then examined who generated the costs and how employees used the tools. Around 10% to 15% of workers produced about 60% of total AI spending. One engineer used $50,000 worth of AI tokens in one month. Employees also often selected costly frontier models for routine tasks.
Management launched an urgent cost review rather than ending employee access. The company sought spending caps across its AI providers and studied the return from each use. Its review covered prompt activity, employee roles, engineering output and code review requests. These findings shaped AI Spend Console’s measurement system and gateway design during 2026.
Rippling negotiated spending limits with Cursor, OpenAI, and Anthropic. It also built an AI gateway that routes each request to a suitable model. The system considers the task and model cost before assigning the work. Companies may connect another gateway, but Rippling requires its gateway for spending controls.
The changes reduced token spending from 40% of Rippling’s R&D headcount budget to about 15%. Usage stayed near its earlier peak. Rippling used 605 billion tokens in April and 600 billion in July. However, July’s token bill equaled 37% of April’s cost, according to MacInnis.
Rippling also compared models through internal tests. CEO Parker Conrad said GLM 5.2 cost 85% less while producing nearly comparable results for the company’s tasks. The company now selects models at several price points instead of using one costly option for every request.
AI Spend Console combines prompt volume, model costs, and work data in one dashboard. For engineers, it can compare spending with code output, pull requests, and review results. Managers can identify employees whose costly AI work often requires revision.
The company also appointed effective AI users as “AI captains.” Those employees help colleagues apply the tools to suitable tasks. Software engineers currently account for most internal usage. Rippling is testing broader uses in customer onboarding, including mail data and reconciliation work.
For onboarding teams, the dashboard can measure output through the number of customers processed. MacInnis said Rippling must connect token use in administrative and customer-facing roles with productivity. Otherwise, the company may restrict access for roles where it cannot measure results.
AI Spend Console comes with Rippling’s HR subscriptions, although customers pay added usage-based AI fees. Companies may also buy it separately and connect it to another HR system. The product tracks individual, team, and role-level spending within the same workforce data structure.
