Following Massive AI Investments, Rippling Launches Innovative Employee ROI Solution

This week, software provider Rippling introduced the AI Spend Console, designed to help organizations monitor and manage their expenditures on AI technologies. A standout feature of this platform is its ability to analyze how individual employees, teams, and roles are spending money on AI and whether their productivity justifies these costs.

The firm asserts that the tool will reveal which engineers are incurring significant AI costs and whether their coworkers frequently request revisions during code evaluations.

The creation of this tool followed Rippling’s initial enthusiasm for AI spending at the beginning of the year, leading to unexpected expenses. Chief Product Officer Matt MacInnis remembers a meeting in March when CFO Adam Swiecicki shared a startling statistic.

Rippling was projected to spend 40% of its R&D budget on AI tokens, equating to a similar amount as the total compensation for 40% of its employees in that department, amounting to millions of dollars. At that time, expenses were surging at a rate of 80% each month, suggesting that if this continued, AI token expenditures would reach 90% of what was allocated for high-salaried R&D personnel the following year.

“We were stunned,” MacInnis commented.

Faced with such alarming data, management promptly initiated a critical assessment of the spending and its effectiveness. The promotional material for the new product humorously depicts Swiecicki sitting on a stool while employees toss large sums of cash into a shredder.

Through its investigation, Rippling identified that about 10-15% of employees were responsible for approximately 60% of total AI-related expenses, with one engineer reportedly spending $50,000 monthly on AI services.

The aim was not to eliminate AI usage but to significantly reduce it. The company began negotiating spending limits with its providers, including Cursor, OpenAI, and Anthropic, only to find that employees tended to default to the latest and most expensive AI models for all tasks.

“The reality is that inference providers like Anthropic and OpenAI have no reason to help you control costs. Their incentive is for expenses to spiral out of control, which is exactly what has happened. They lack transparency on usage and do not collaborate effectively,” said MacInnis.

This scenario has shifted over the past eight months, as companies have learned crucial lessons. They now realize the importance of using multiple AI models from various labs that cater to different budget levels, including potentially lower-cost options.

Last month, Rippling’s CEO Parker Conrad stated that internal benchmarks indicated SpaceX’s Grok performed exceptionally well, whereas GLM 5.2, which is 85% cheaper, yielded nearly identical outcomes. Demand for GLM 5.2 has grown among tech firms for programming tasks, particularly after endorsements from Databricks.

Additionally, companies understand the need for an AI gateway that directs requests to the most suitable and economical model. Rippling took this insight further by developing its own AI gateway, which is integrated into the AI Spend Console. Existing users of other gateways can still utilize this product, but will need to switch to Rippling’s for enhanced spending management features.

The AI Spend Console generates dashboards that evaluate metrics like daily prompts alongside productivity metrics (e.g., lines of code or pull requests) and related spending.

With this system in place, Rippling reduced its token expenditures from 40% of its headcount budget to 15%. Notably, while the internal token usage reached 600 billion tokens in July, which mirrored a peak from April, the overall expense for July’s tokens was only 37% of April’s total costs, attributed to more efficient model routing.

“Now we’re directing tasks to more effective models,” he added, humorously noting that “we’re not letting the sales team edit grammar using Fable.”

However, technological measures alone are not sufficient, Rippling points out. The company has identified employees who effectively use AI and designated them as “AI captains” to mentor others.

Despite progress, efforts to expand AI usage beyond engineering teams remain ongoing, MacInnis noted. For instance, Rippling is exploring applications for customer onboarding teams to automate specific tasks, with the dashboard focusing on onboarding efficiency metrics.

“We must connect token usage in general and customer-facing activities to productivity. If we can’t achieve this, we may have to restrict access to all employees,” MacInnis cautions.

The AI Spend Console is available to Rippling’s HR subscribers, though additional fees apply based on AI usage. It can also be bought separately and integrated with other HR systems.