Rippling Launches AI Spend Tool After Millions Spent on AI Tokens

Rippling has launched AI Spend Console, a new product designed to help companies understand how much they are spending on AI and whether that spending is actually making employees more productive.

The HR software provider says the product was born out of its own experience with “tokenmaxxing,” after the company went all in on AI earlier this year and discovered that employees were burning through millions of dollars in AI usage.

Matt MacInnis, Rippling’s chief product officer, recalled an executive meeting in March when CFO Adam Swiecicki presented a number that shocked the leadership team: The company was on track to spend the equivalent of 40% of its R&D headcount budget on AI tokens.

AI spending was growing by roughly 80% month over month. If that trend had continued, Rippling calculated that within a year it could have been spending almost as much on AI tokens as it spent compensating its entire R&D workforce.

“We were incredulous,” MacInnis said.

The company immediately began an internal analysis to determine where the money was going and what Rippling was getting in return. It found that roughly 10% to 15% of employees were responsible for about 60% of total AI spending. One engineer alone was spending approximately $50,000 a month.

Rippling wanted to control spending, not AI usage

Rippling did not want to stop employees from using AI. Instead, it wanted to make sure workers were using the right models for the right tasks, and that expensive AI usage was translating into useful work.

The company began by negotiating maximum spending limits with the AI tools it used, including Cursor, OpenAI and Anthropic. It also quickly discovered that employees tended to default to the newest and most expensive frontier models, even when less expensive systems could handle many tasks just as effectively.

MacInnis argued that AI providers have little incentive to help companies control their spending because their businesses benefit when customers consume more tokens.

“The truth is that the inference providers, like Anthropic and OpenAI, have absolutely no incentives to help you control your spend,” MacInnis said. “They have every incentive for it to be a runaway expense.”

Rippling eventually concluded that enterprises needed access to multiple models at different price points, including frontier and open-weight systems. Its own internal testing found that SpaceX’s Grok performed strongly across a range of tasks, while Z.ai’s GLM 5.2 delivered nearly identical performance at about 85% lower cost for some workloads.

That led Rippling to build an AI gateway that automatically routes prompts to the model that makes the most sense for a particular task. The gateway is now part of AI Spend Console, although companies already using another AI gateway can still use parts of Rippling’s product.

From token spending to employee ROI

AI Spend Console provides dashboards that Rippling says can show how much individual employees, teams and roles are spending on AI and whether that spending is translating into measurable output.

The system can compare factors such as prompts per day, AI spending and work output, including lines of code and pull requests. Rippling says the goal is to identify not only its biggest AI users, but whether those users are actually producing better work.

The company says its changes have already reduced AI spending from the equivalent of 40% of its R&D headcount budget to about 15% without meaningfully cutting usage.

Rippling consumed a peak of roughly 605 billion tokens in the month when its CFO raised the alarm. In July, internal usage reached about 600 billion tokens again, but the cost of that usage was only 37% of what the company paid in April.

MacInnis attributed the difference largely to better model selection and routing. Instead of sending every request to an expensive frontier model, Rippling can now direct simpler workloads to cheaper systems.

Measuring AI beyond engineering

Rippling says controlling AI costs also requires people to help their colleagues use the technology effectively. The company has identified employees who are particularly successful with AI and designated them as “AI captains” who can help others adopt better workflows.

Software engineers remain the company’s biggest AI users, but Rippling is beginning to apply the same approach to other departments. For example, it is testing AI for customer onboarding tasks involving mailing data and data reconciliation.

In those cases, the company wants to measure productivity in terms of business outcomes, such as how many customers can be onboarded. The broader goal is to connect token consumption in general and administrative and customer-facing functions directly to measurable productivity.

“We have to be able to link token consumption in G&A functions and in customer-facing functions back to productivity,” MacInnis said. “If we can’t do that, all bets are off on any of this stuff being available to the broader employee base.”

That could change how companies think about workplace AI. Rather than treating AI access like Slack or email and giving every employee unlimited access, companies may increasingly require evidence that AI spending is producing measurable value.

AI Spend Console is included with Rippling’s HR subscriptions, although additional AI usage-based costs apply. MacInnis said it can also be purchased as a standalone product and integrated with another HR system of record.