HR software company Rippling just launched AI Spend Console, a product built to track and rein in how much a business burns on AI. According to TechCrunch AI, the tool grew out of Rippling’s own painful discovery earlier this year: employees were torching cash on AI tokens at a rate nobody saw coming. What makes this launch interesting is that Rippling isn’t selling theory. It’s selling the fix it built to save itself.
The origin story is the sell. Chief Product Officer Matt MacInnis told TechCrunch AI about a March executive meeting where the CFO revealed the company was on track to spend 40% of its entire R&D headcount budget on AI tokens. Millions of dollars. Spending was climbing 80% month over month. “We were incredulous,” MacInnis said. When Rippling dug in, it found that 10 to 15% of employees drove roughly 60% of total AI spend, and one engineer was burning $50,000 a month.
What AI Spend Console actually does
The product is aimed squarely at what MacInnis calls “tokenmaxxing,” the early-2026 habit of throwing the newest, priciest frontier model at every task. Key features, per TechCrunch AI:
- Per-employee spend mapping. It breaks down AI costs by individual, team, and role, so finance can see exactly where the money goes.
- Productivity scoring. Dashboards combine prompts per day with real output like lines of code and pull requests, then weigh that against spend.
- Slop detection. The blog post promises it will flag “which engineers have high AI spend whose peers frequently ask them to redo work in code reviews.”
- A built-in AI gateway. It routes each prompt to the most cost-effective model for the job instead of defaulting to the most expensive one.
That last piece matters. MacInnis was blunt about why companies got burned: “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.”
The results Rippling is pointing to
Rippling used its own tool internally and cut token spend from 40% of its headcount budget down to about 15%. Here’s the part that stands out: it didn’t slash AI usage to get there. The company hit a peak of 605 billion tokens in the month of the CFO’s warning, then hit 600 billion again in July. Yet July’s token cost was just 37% of April’s. The difference was smart routing. As MacInnis put it, “we’re not letting the sales team do grammar updates using Fable.”
Rippling’s benchmarks pushed it toward a multi-model setup. CEO Parker Conrad found that SpaceX’s Grok led overall, but that Z.ai’s GLM 5.2 was “85% cheaper but nearly identical” in performance. That Chinese coding model has become a quiet favorite among tech companies, with Databricks also backing it.
Availability and the catch
AI Spend Console is included for Rippling’s existing HR subscribers, though there are extra AI usage-based costs on top. Companies on a different HR system can buy it as a standalone product and integrate it. One caveat worth flagging: if you want the features that actually govern spending, you have to use Rippling’s gateway, even if you already run one of your own.
There’s also a cultural shift baked in. Rippling identified its most effective AI users and named them “AI captains” to coach everyone else. And the company is honest that this works best for engineers so far. Extending it to onboarding and customer-facing teams is still a work in progress, because those functions are harder to tie back to measurable output.
That points to something bigger. MacInnis says that if a company can’t link token spend to productivity in a given function, “all bets are off” on giving those employees AI access at all. In other words, the tokenmaxxing era may have swung so far the other way that broad AI access is no longer a given like Slack or email. Prove the productivity, or lose the tokens.
Watch whether this becomes a category. If enterprises really have moved from “give everyone AI” to “justify every token,” tools that measure ROI per employee could get crowded fast. More details are available in the original TechCrunch AI report.