Rippling Was Burning Millions on AI. So It Built a Tool to Stop It.
The HR software company discovered one engineer was spending $50,000 a month on AI tokens. Its new AI Spend Console is designed to make sure that never happens to you.
Key points
- Rippling's AI token spending hit 40% of its entire R&D staff compensation budget before the company intervened in early 2025.
- A single engineer at Rippling was spending $50,000 a month on AI tokens, the company confirmed.
- Roughly 10-15% of Rippling's employees were responsible for about 60% of its total AI spending.
- After deploying its own AI routing tool, Rippling cut token costs to 37% of what it paid in April, while using nearly the same number of tokens.
- AI Spend Console is included for existing Rippling HR subscribers, with additional usage-based costs, and can also be bought as a standalone product.
Rippling, the company best known for HR and payroll software, has launched a new product called AI Spend Console. The short version: it tells businesses exactly who on their team is spending what on AI tools, and whether they have anything to show for it.
The slightly longer version involves a CFO, a paper shredder, and millions of dollars.
What went wrong at Rippling?
Early this year, Rippling went all-in on AI tools for its engineers. Costs spiralled fast. By March, Chief Financial Officer Adam Swiecicki presented a figure in an executive meeting that stopped the room cold.
Rippling was on track to spend the equivalent of 40% of its entire R&D (research and development, meaning the engineering team) salary budget on AI tokens. Tokens are the small units of text that AI services charge for, roughly like paying per word processed. Spending was growing at 80% every month.
If that trend continued, token costs would equal 90% of what Rippling paid its engineers within a year.
"We were incredulous," Chief Product Officer Matt MacInnis told TechCrunch AI.
An internal audit turned up some stark numbers. One engineer alone was spending $50,000 a month. Around 10-15% of staff were driving roughly 60% of total spending. Part of the problem was simple: employees defaulted to the newest, most expensive AI models for every task, including ones a cheaper model could handle just as well.
MacInnis put it plainly: "The inference providers, like Anthropic and OpenAI, have absolutely no incentives to help you control your spend. They have every incentive for it to be a runaway expense."
What does AI Spend Console actually do?
It tracks AI spending by individual, team, and role, and pairs that spending with real output. For engineers, that means looking at pull requests (bundles of code changes submitted for review) and lines of code written alongside what they spent. The tool can flag, for example, which engineers spend heavily on AI but whose colleagues regularly ask them to redo their work.
Rippling also built its own AI gateway, a routing layer that automatically sends each task to the most cost-effective model suitable for the job, rather than always defaulting to the priciest option. The company found that models vary a lot by task. Parker Conrad, Rippling's CEO, noted last month that for internal coding work, Z.ai's GLM 5.2 model was 85% cheaper than frontier (top-tier, most capable) models with nearly identical results.
| Month | Tokens used | Cost vs April |
|---|---|---|
| April 2025 (peak) | 605 billion | 100% (baseline) |
| July 2025 | ~600 billion | 37% of April cost |
| Budget share (before) | 40% of R&D salaries | , |
| Budget share (after) | ~15% of R&D salaries | , |
Rippling used roughly the same volume of tokens in July as it did in April. The bill was 63% lower. MacInnis joked they stopped "letting the sales team do grammar updates using" top-tier models built for complex reasoning.
Should your company care about this?
If your business pays for AI tools and has more than a handful of employees using them, yes. The pattern Rippling hit is not unusual. Tools like Cursor (a coding assistant), Claude from Anthropic, and ChatGPT from OpenAI all charge per use, and costs add up quietly.
AI Spend Console is free for existing Rippling HR customers, though usage-based charges apply on top. It is also available as a standalone product that connects to other HR systems. Rippling has not published standalone pricing publicly.
One practical note: the productivity tracking features require you to use Rippling's own AI gateway. If your company already uses a different routing tool, the dashboards still work, but the spending controls need Rippling's gateway to function.
Common questions
What is an AI token, and why does it cost money?
A token is a small chunk of text, roughly three-quarters of a word, that AI services process when you send or receive a message. Companies like OpenAI and Anthropic charge per thousand tokens, so a team sending thousands of long prompts every day can run up a surprisingly large bill.
Does this tool restrict what employees can do with AI?
Rippling says the goal is smarter routing, not a blanket block. Employees keep access to AI tools; the system steers cheaper models to routine tasks and reserves expensive ones for work that genuinely needs them.
Is my data private if Rippling monitors AI usage?
The Console tracks spending and output metrics inside your company's Rippling account. Your prompts route through Rippling's gateway, which means Rippling sees metadata about usage volumes and costs. If your team handles sensitive data in AI prompts, check Rippling's data-processing terms before deploying.
