Rippling's AI Spend Console: What Indian Enterprises Can Learn About AI Leadership
Here's a number that should keep every C-suite executive up at night: Rippling was on track to burn 40% of its R&D headcount budget on AI tokens in March 2026. Forty percent. That's not a rounding error. That's spending as much on AI tokens as the company paid in compensation to 40% of its engineering workforce — millions of dollars, bleeding out monthly, with spending growing 80% month-over-month.
That's the kind of trajectory that gets you fired, or worse, bankrupt.
Rippling, the HR software provider founded by CEO Parker Conrad, didn't panic. It didn't pull the plug on AI either. Instead, it did something most companies are still fumbling with: it built an actual system to track who's using AI, how much they're spending, and whether that spending is actually making anyone more productive. The result, unveiled August 7, 2026, is called the AI Spend Console. And it's worth paying attention to — especially if you're an Indian enterprise leader trying to figure out how to lead your own AI transformation without setting fire to your budget.
The Tokenmaxxing Problem Nobody Saw Coming
Rippling went all in on AI at the start of 2026, like so many other companies. Employees got access to Cursor, OpenAI, Anthropic — the whole frontier toolkit. Nobody set spending caps. Nobody thought about it. The assumption was simple: more AI access equals more productivity.
The assumption was wrong.
Chief Product Officer Matt MacInnis recalls the March executive meeting when CFO Adam Swiecicki presented the numbers. The room went quiet. Roughly 10–15% of Rippling's employees were driving about 60% of total AI spend. One engineer alone was burning $50,000 a month on AI tokens. Not $5,000. Not $500. Fifty thousand dollars. Monthly.
And here's the thing that made it worse: employees defaulted to using the most expensive frontier models for every single task. Even grammar checks. Even simple code reviews. They were using GPT-4-class models to do what a cheap model could handle, and nobody had noticed until the bill arrived.
"We were incredulous," MacInnis told TechCrunch.
The launch ad for the AI Spend Console actually features Swiecicki sitting on a stool while employees pick up wads of cash and dump them into a paper shredder. It's darkly funny, but the underlying message is serious: companies were hemorrhaging money on AI without understanding what they were getting for it.
Building an Anti-Tokenmaxxing Playbook
Rippling didn't want to stop AI usage. They wanted to rein it in — a lot. Their approach had three parts, and each one is worth studying.
First, they negotiated maximum spending caps with each tool provider: Cursor, OpenAI, and Anthropic. This was the low-hanging fruit. Second, they built their own AI gateway, software that routes prompts to the best, most cost-effective model for each specific task. Third, they built dashboards that actually score employees on prompts per day combined with work output (lines of code, pull requests) and spend.
MacInnis makes a sharp observation about the current AI ecosystem: "The truth is that 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, and that's exactly what they do. They don't provide you with great usage insight, and they don't collaborate with one another."
That's a fair critique. The AI providers are businesses. They make money when you use more tokens. They're not your cost consultants.
Conrad, Rippling's CEO, took it a step further by benchmarking models for internal use. SpaceX's Grok came out ahead on benchmarks, but here's the kicker: Z.ai's GLM 5.2 was 85% cheaper with nearly identical performance to the frontier models. GLM 5.2 has become a particular favorite for coding tasks among tech companies, with Databricks also championing it. The lesson? You don't always need the most expensive model. Sometimes you just need the right model for the job.
The Numbers That Actually Matter
The results speak for themselves. After implementing the AI Spend Console, Rippling dropped its token spend from 40% of its headcount budget to about 15%. That's a massive reduction. But here's what makes it interesting: they didn't reduce AI usage. The company spent a peak of 605 billion tokens in April, the month Swiecicki issued his warning. In July, internal usage hit 600 billion tokens again, essentially the same volume.
"The cost of July's token spend was 37% of the cost of April's token spend," MacInnis said. "That's just because now we're routing to the more effective models."
He joked that "we're not letting the sales team do grammar updates using Fable," but the point stands. By routing prompts to cheaper, more appropriate models, they cut costs dramatically while maintaining output.
The dashboards, once called leaderboards during the tokenmaxxing era, now score attributes like prompts per day, work output, and spend. It's a system that rewards productivity, not just consumption.
Beyond Engineering: The Hard Part
Technology solutions aren't enough, Rippling found. The company identified employees who used AI effectively and made them "AI captains" tasked with helping the rest of the organization. That's a smart organizational move, peer-to-peer knowledge transfer tends to work better than top-down mandates.
But the real challenge lies beyond engineering. Software engineers have been the primary AI users so far, and for good reason: AI tools integrate naturally into coding workflows. The harder work is extending AI productivity gains to other functions.
Rippling is currently working on applying the AI Spend Console to customer onboarding teams, automating mailing data and data-reconciliation tasks. The dashboard would then measure productivity in terms of onboarding more customers, not lines of code.
"We have to be able to link token consumption in G&A functions and in customer-facing functions back to productivity," MacInnis says. "If we can't do that, all bets are off on any of this stuff being available to the broader employee base."
That's the rub, isn't it? If companies can't measure productivity, they might just pull AI access entirely. And that would be a mistake, a waste of a genuinely transformative technology.
What This Means for Indian Enterprises on AI Leadership
For Indian enterprises navigating their own AI transformations, Rippling's experience offers several lessons. First, don't wait for the bill to shock you. Set spending caps early. Second, build an AI gateway rather than relying on individual tool providers, they're not designed to help you control costs. Third, benchmark models rigorously. The cheapest model isn't always the worst performer, and the most expensive isn't always the best.
The AI Spend Console is included for Rippling's HR subscribers, with additional AI usage-based costs. It can also be purchased as a standalone product and integrated with another HR system of record, according to MacInnis. That matters for enterprises that already use different HR platforms.
Tokenmaxxing may have swung so far the other direction that employee AI access might not become like Slack or email, ubiquitous, untracked, unmeasured. If companies can't measure productivity, they might restrict access entirely. Rippling's approach suggests a middle path: measure everything, reward productivity, and route intelligently.
For Indian enterprises looking to lead their own AI transformations, that's a playbook worth studying. The question isn't whether to use AI. It's whether you'll spend millions figuring it out the hard way, or learn from someone who already did.
