Rather than shutting off access, Rippling built AI Spend Console, launching it publicly on August 6, 2026 . The product has several interconnected capabilities:

Per-employee spend tracking

AI Spend Console maps AI token spending to individual employees, teams, and roles, then correlates it against productivity signals such as pull requests and code reviews . This gave Rippling the ability to see which specific roles or individuals were driving the bill — and whether that spend was producing results.

AI gateway with smart routing

Rippling built an intermediary layer that automatically routes each prompt to the most cost-effective model for the task instead of defaulting to expensive frontier models . The company's internal benchmarks found that Z.ai's GLM 5.2 model was 85% cheaper than leading frontier models while delivering nearly identical performance, so routine prompts were redirected there . (Note: this is a company-reported comparison, not an independent result .)

Spend caps per tool

Rippling negotiated a maximum spending cap for each AI tool — Cursor, OpenAI, Copilot, and others — to prevent any single tool from running up an uncapped bill .

Productivity correlation

The console ties token consumption to measurable output — code commits, PR quality, review velocity — so that spend that doesn't yield productivity can be flagged and acted on . Rippling pointedly calls it a "dashboard" rather than a "leaderboard" .

AI captains

Designated employees responsible for monitoring and governing AI spend within their teams, creating direct accountability .

The result

Within months, Rippling cut its AI token spend from ~40% of R&D headcount budget to roughly 15% — without reducing usage volume. The company was still consuming approximately 600 billion tokens per month in July, but the July bill was only 37% of what it was in April . As multiple sources put it: the token volume never fell — the bill did .

Rather than shutting off access, Rippling built AI Spend Console, launching it publicly on August 6, 2026 . The product has several interconnected capabilities:

Per-employee spend tracking

AI Spend Console maps AI token spending to individual employees, teams, and roles, then correlates it against productivity signals such as pull requests and code reviews . This gave Rippling the ability to see which specific roles or individuals were driving the bill — and whether that spend was producing results.

AI gateway with smart routing

Rippling built an intermediary layer that automatically routes each prompt to the most cost-effective model for the task instead of defaulting to expensive frontier models . The company's internal benchmarks found that Z.ai's GLM 5.2 model was 85% cheaper than leading frontier models while delivering nearly identical performance, so routine prompts were redirected there . (Note: this is a company-reported comparison, not an independent result .)

Spend caps per tool

Rippling negotiated a maximum spending cap for each AI tool — Cursor, OpenAI, Copilot, and others — to prevent any single tool from running up an uncapped bill .

Productivity correlation

The console ties token consumption to measurable output — code commits, PR quality, review velocity — so that spend that doesn't yield productivity can be flagged and acted on . Rippling pointedly calls it a "dashboard" rather than a "leaderboard" .

AI captains

Designated employees responsible for monitoring and governing AI spend within their teams, creating direct accountability .

The result

Within months, Rippling cut its AI token spend from ~40% of R&D headcount budget to roughly 15% — without reducing usage volume. The company was still consuming approximately 600 billion tokens per month in July, but the July bill was only 37% of what it was in April . As multiple sources put it: the token volume never fell — the bill did .