HR software vendor Rippling this week launched AI Spend Console, a product designed to track and contain employee AI-token spending. The tool goes beyond aggregate spend reporting: it maps token consumption by individual employees, teams and roles and attempts to connect that spending to measurable productivity rather than simply showing raw usage.
Why the tool was developed
The product grew out of Rippling’s own experience earlier in 2026. After broadly enabling high-volume use of large models at the start of the year, company executives were alarmed in March when CFO Adam Swiecicki presented internal figures. Rippling found it was on track to spend roughly 40% of its R&D headcount budget on AI tokens — effectively allocating as much to tokens as to 40% of the unit’s total compensation. Month-over-month token spending had been growing at about 80%, and if that trend had continued the next year it could have approached roughly 90% of the R&D compensation cost.
An internal analysis revealed further details: about 10–15% of employees accounted for roughly 60% of total AI spend, and one engineer was spending about $50,000 per month. At the peak month referenced by the CFO, internal usage reached 605 billion tokens.
What Rippling did in response
Rippling did not aim to ban AI use, but to control it. The company first negotiated maximum spending caps with the providers it used, including Cursor, OpenAI and Anthropic. That process exposed a common problem: employees tended to default to the newest and most expensive frontier models for all tasks.
Rippling executives noted that inference providers such as Anthropic and OpenAI have limited incentives to help customers control costs; the providers benefit when usage is high, and they generally do not offer detailed, coordinated usage insights.
Industry lessons in 2026
The company’s experience reflects wider industry lessons from early 2026. Enterprises have increasingly recognized the need for multiple models at different price points — including cheaper, open-weight frontier options. Rippling’s internal benchmarks, cited by founder and CEO Parker Conrad, suggested that SpaceX’s Grok performed strongly overall but that GLM 5.2 (from Z.ai) was about 85% cheaper with nearly identical performance for many tasks. (SpaceX now owns Cursor, which provides access to Grok and other models.) Databricks has also promoted such models.
Another key conclusion is the value of an AI gateway that routes prompts to the most appropriate and cost-effective model. Rippling built its own gateway as part of AI Spend Console; the company says organizations using another gateway can still adopt the console, but to use Rippling’s spending-governance features they would need Rippling’s gateway.
Features and outcomes
AI Spend Console produces dashboards (previously described as leaderboards during early tokenmaxxing) that score metrics like prompts per day combined with work output (lines of code, pull requests) and spend. According to Rippling, after deploying the tool they reduced token spend from about 40% of their R&D headcount budget to roughly 15% while retaining AI usage.
The company reported that internal usage again reached roughly 600 billion tokens in July, but the cost of July’s token spend was only 37% of the cost incurred in April — a result attributed to routing requests to more cost-effective models.
People and process changes
Rippling emphasizes that tooling alone is not enough. The company identified employees who used AI effectively and appointed them as “AI captains” to help others adopt better practices. So far, software engineers have been the primary users, and the company is still working to expand productive AI use into functions such as customer onboarding, where the dashboard would measure productivity in terms of onboarding more customers tied to token consumption.
Availability and pricing
AI Spend Console is included for Rippling HR subscribers, though AI usage incurs additional, usage-based costs. The product can also be bought standalone and integrated with another HR system of record, Rippling says.
Conclusion
Rippling’s experience illustrates how unrestricted token use can rapidly drive costs, and how combining model choice, routing, governance and organizational support can substantially improve cost-efficiency while preserving AI access. The company also warns that if organizations cannot connect token consumption to productivity, broad employee access to AI may need to be limited.
