AI has quickly become a major expense for companies, and HR software company Rippling learned that lesson the hard way. After seeing its internal AI bill grow dramatically, the company has now introduced a new product designed to help businesses understand whether their AI spending is actually improving productivity.
Called AI Spend Console, the tool gives companies a detailed view of how much they are spending on AI across employees, teams, and job functions. More importantly, it attempts to connect that spending with measurable results.
For example, companies can identify engineers with unusually high AI usage and compare that activity with indicators such as code output and the amount of rework required during reviews.
A costly wake-up call
Rippling’s new system came from an internal problem. Earlier this year, executives discovered that AI token costs were consuming an amount equivalent to roughly 40% of the company’s R&D compensation budget.
The expense was also increasing by around 80% every month. If that pace had continued, AI tokens could eventually have represented an enormous portion of what Rippling spent on its engineering organization.
An internal review revealed that only about 10% to 15% of employees were responsible for approximately 60% of the company’s AI expenditure. One engineer alone was reportedly using around $50,000 worth of AI services every month.
Rather than banning AI tools, Rippling decided to understand and control how employees were using them.
Smarter model selection
The company initially negotiated spending limits with providers including OpenAI, Anthropic, and Cursor. It soon discovered another problem: employees frequently selected the newest and most expensive models, even when cheaper alternatives could handle the same tasks.
Rippling responded by developing its own AI gateway. The system can route requests toward different models depending on the job, balancing performance and cost.
That approach produced significant savings. The company reduced AI spending from the equivalent of 40% of its R&D headcount budget to approximately 15%, while maintaining heavy usage.
At one point, employees consumed around 605 billion tokens in a month. Later, usage reached approximately 600 billion tokens again, but the cost was dramatically lower because more affordable models were being used for suitable workloads.
Measuring productivity
AI Spend Console combines usage information with productivity indicators, creating dashboards that show companies where AI is being used and whether the spending appears justified.
Rippling has also appointed employees as “AI captains” to help colleagues use these tools more effectively.
For now, software engineering remains the biggest area of AI adoption. However, the company is experimenting with applications in other departments, including customer onboarding, where AI can help automate data processing and reconciliation.
The larger question is whether companies can prove that AI spending produces measurable business value. Rippling’s experience suggests that simply giving every employee unlimited access may not be sustainable.
AI Spend Console is included with Rippling’s HR platform, although AI usage can involve additional costs. The company also offers the product separately for organizations using other HR systems.
