Rippling Cuts AI Spending by 63% After Burning Millions on Tokens, Launches Control Tool

Rippling Faces AI Spending Crisis

HR software provider Rippling confronted a startling reality in March when CFO Adam Swiecicki revealed the company was burning through millions on AI tokens at an alarming rate. The firm was on track to spend 40% of its R&D headcount budget on AI tokens, with costs escalating 80% month-over-month1

1

. If the trend continued, Rippling would spend nearly 90% of what it paid R&D employees on tokens alone within a year. Chief Product Officer Matt MacInnis recalled the executive team's shock: "We were incredulous." Analysis uncovered that roughly 10-15% of employees drove 60% of total AI spending, with one engineer alone burning $50,000 monthly1

1

.

AI Spend Console Emerges from Crisis

Rippling responded by developing AI Spend Console, a product designed to help companies manage and optimize AI spending while tracking productivity outcomes. The tool maps individual employee, team, and role-level AI token consumption against actual work output. For engineers, it correlates spending with lines of code, pull requests, and code review results, identifying "which engineers have high AI spend whose peers frequently ask them to redo work in code reviews"1

1

. The dashboard combines prompt volume, model costs, and workforce data to reveal whether employees genuinely boost productivity or generate low-quality output2

2

.

Multi-Model Approach Delivers Massive Savings

Rippling discovered employees defaulted to expensive frontier models for every task, regardless of complexity. MacInnis criticized AI providers: "The truth is that the inference providers, like Anthropic and OpenAI, have absolutely no incentives to help you control your spend"1

1

. The company implemented a multi-model approach, negotiating spending caps with Cursor, OpenAI, and Anthropic while building an AI gateway to route prompts to cost-effective models. Parker Conrad noted that internal benchmarks found SpaceX's Grok performed best overall, but GLM 5.2 delivered 85% cost savings with nearly identical performance for Rippling's specific tasks1

1

2

2

. The Chinese-origin GLM 5.2 has become particularly popular for coding tasks among tech companies.

Dramatic Cost Reduction Without Usage Decline

After implementing AI Spend Console and its gateway, Rippling reduced token spend from 40% to 15% of its R&D headcount budget1

1

2

2

. The company consumed 605 billion tokens in April at peak spending, then hit 600 billion tokens again in July, yet July's costs were just 37% of April's expenses. MacInnis attributed the savings to routing requests to more effective models, joking they stopped "letting the sales team do grammar updates using Fable"1

1

. The results demonstrate companies can maintain AI usage levels while dramatically cutting costs through intelligent model selection.

AI Captains and Expanding Use Cases

Rippling appointed effective AI users as "AI captains" to help colleagues apply tools appropriately1

1

2

2

. While software engineers remain primary users, the company is testing broader applications in customer onboarding for automating mail data and reconciliation tasks1

1

2

2

. For onboarding teams, dashboards will measure productivity through customers processed. MacInnis acknowledged the challenge of connecting token use in administrative roles with measurable results, warning Rippling may restrict access where productivity cannot be tracked. AI Spend Console comes bundled with Rippling's HR subscriptions, though customers pay additional usage-based AI fees, and companies can purchase it separately to connect with other HR systems2

2

.