Rippling launches AI tool to track employee spending

After its own AI token costs surged, Rippling built AI Spend Console to tie usage to output and help companies cut waste without limiting use.

HR software provider Rippling has introduced AI Spend Console, a new product aimed at helping companies track and control what they spend on AI. The tool grew out of Rippling’s own experience after its internal AI costs climbed sharply and leadership tried to understand whether that spending was actually improving work. For businesses pushing AI deeper into daily operations, that is becoming a more pressing question. Costs can rise fast, while the payoff is often harder to measure.

The launch puts a spotlight on a broader problem in enterprise AI. Spending may look manageable at first, then escalate quickly once teams start using multiple tools and expensive models. Rippling says its console is designed to show where AI is being used heavily, who is using it, and whether that usage is leading to useful output or simply more low-quality work.

Rippling built AI Spend Console after its own costs jumped

The product took shape after Rippling expanded AI use heavily earlier this year and then found the bill was much higher than expected. Chief Product Officer Matt MacInnis said leadership was alarmed during a March meeting when CFO Adam Swiecicki presented figures showing the company was on track to burn 40% of its R&D headcount budget on AI tokens.

In practical terms, that meant spending on AI tokens at a level equal to 40% of what Rippling paid employees in its research and development unit. The company described that as millions of dollars. It also said token spending was rising by 80% month over month, and if that trend had continued, next year’s AI token bill could have reached almost 90% of what it spent on its high-paid R&D staff.

MacInnis said management responded with an urgent effort to understand where the money was going and what the company was getting back. According to IT-PUB News, Rippling’s own analysis found that roughly 10% to 15% of employees were responsible for about 60% of total AI spending. The company also said one engineer was spending $50,000 a month.

The dashboard ties AI usage to work output

AI Spend Console centers on dashboards that score AI usage alongside work output. Rippling says the tool can track prompts per day, lines of code, pull requests, and spending, giving managers a way to see not just how much AI is being used, but whether that usage appears productive.

The system can also flag patterns such as engineers whose peers often ask them to redo work during code reviews while those same engineers are still generating high AI costs. In other words, Rippling wants the product to measure both expense and output quality.

That reflects a wider shift in how enterprises are approaching AI. Instead of simply pushing for more usage, companies are increasingly trying to work out which tasks justify expensive frontier models and which can be handled more cheaply.

Rippling says model routing lowered costs

Rippling says it did not want to cut AI use altogether. Instead, it tried to control spending by negotiating maximum budgets with the AI tools it used, including Cursor, OpenAI and Anthropic.

That effort quickly exposed another issue. Employees were often defaulting to the newest and most expensive frontier models, even for tasks that did not necessarily require them. Rippling argues that AI providers have little incentive to help customers limit spending or compare usage across tools.

MacInnis said the company decided it needed better routing, so it built its own AI gateway as part of the product. The gateway sends prompts to the model Rippling considers the best and most cost-effective for the task. Enterprises that already use another gateway can still use AI Spend Console, he said, but they would need Rippling’s gateway to get the spending controls.

Rippling says the results were significant. The company says it reduced token spending from 40% of its headcount budget to about 15% without reducing AI use. It said internal usage peaked at 605 billion tokens in the month when the CFO raised the alarm. In July, usage again hit 600 billion tokens, but the cost of that month’s token spend was 37% of April’s because more work was being routed to cheaper and more effective models.

MacInnis joked that the company was no longer allowing the sales team to use Fable for grammar edits. The point was straightforward: not every task needs the most expensive AI model.

Rippling is trying to push AI beyond engineering

Even with the new tool, Rippling says software engineers remain the main AI users inside the company. It has also started working with other departments, including customer onboarding teams, where it wants to automate some mailing-data and data-reconciliation tasks.

Rippling says it is also trying to measure productivity in those non-engineering areas, not just token use. MacInnis said the company needs to connect AI consumption in general and administrative functions, as well as customer-facing roles, to actual productivity. If it cannot do that, he said, the company cannot confidently make AI widely available across the workforce.

The company also said it has identified employees who use AI effectively and turned them into “AI captains” to help others across the business.

That gets at a larger issue for employers. AI adoption is no longer just about buying access to tools. It also means deciding who gets access, how much they can use, and whether the company can show the spending is worth it.

Rippling is now selling the product to other companies

AI Spend Console is included for Rippling’s HR subscribers, though the company says additional AI usage-based costs apply. It can also be purchased separately and connected to another HR system of record.

For businesses, the pitch is clear. AI can be useful, but it can also become a runaway expense if companies do not know which teams are using it, for what purpose, and at what cost.

Rippling’s experience suggests the next phase of enterprise AI may be less about giving everyone broad access and more about managing that access carefully. If companies cannot show that AI is improving work, they may grow more cautious about expanding it beyond a smaller group of users.