Larridin research shows that most of the productivity boost from AI coding assistants comes from just one-third of total spending.

Recently, AI "sticker shock" has hit the enterprise. Beginning in mid-May, sudden increases in AI costs alarmed companies—and made news—worldwide:

• More! More! More! Tech Workers Max Out Their A.I. Use, New York Times, 5/20/2026
• The AI Token Pricing Crisis Behind OpenAI and Anthropic’s Revenue Race, Investing.com, 5/22/2026
• Uber burned through its entire 2026 AI budget in four months. Now its COO is questioning whether it’s worth it, Fortune, 5/26/2026
• AI sticker shock hits corporate America, Axios, 5/28/2026
• Client Accidentally Burns $500 Million on Claude AI in One Month: Here’s How, Yahoo! Finance, 5/29/2026
• Anthropic Files to Go Public, Setting Stage for Huge I.P.O., New York Times, 6/1/2026
• CNN Interviews Larridin CEO for Anthropic IPO, from CNN, 6/3/2026

The US leads in private-company AI spend, 2013-2025. Figure courtesy Statista.com

Several trends have contributed to this cost shock, each pushing the alarm meter upwards:

• AI providers introduced, and rapidly improved, agentic AI, which burns tokens at scale, and around the clock. Providers also moved to rigorous per-token payment plans.
• Many AI users, especially software developers, moved quickly (and often, effectively) to use agentic workflows to boost their productivity.
• An employee's token burn rate became a proxy for their effective use of AI. Some companies created token usage leaderboards, promoting heavy AI use, with no tie to productivity.

The result, which seems predictable in hindsight, was an explosion in AI costs in the enterprise.

Is All This Spending Productive?

High spending can be worthwhile, if it results in high productivity. Here at Larridin, we measure AI adoption and helps enterprise customers optimize AI-powered productivity. We have seen our customers use the Larridin platform achieve great, rapid, and well-managed success with AI-powered tools of all kinds.

However, we also observed some of the above trends taking shapte. So Larridin co-founder and CTO Ameya Kanitkar used proprietary data to analyze spending and productivity for AI-powered coding tools

The direction of the findings fit our expectations, but their degree surprised even us:

• Less than a third of users are responsible for more than 50% of spend. The tokenmaxxers, the people at the top of the leaderboards, are of course also leading the way in costs.
• The first one-third of spending is responsible for 85% of productivity gains. The first third or so of spending delivers strong return on investment (ROI). But spending rises much faster than productivity after that point.

Our customers, who see these trends in their Larridin dashboards, are already adjusting their AI usage and workflows to match. The AI companies will be held accountable for whether their customers can achieve sustainable returns on their fast-growing AI investment.

Don't Throw the AI Baby Out with the Bathwater

The problem isn't that AI is ineffective. Used properly, it's surprisingly effective. Our customers' success reminds us of this every day.

And the problem isn't the employees. In many cases, they've been encouraged, subtly or overtly, to maximize their use of tokens, as a way of showing that they're fully up to speed in the new technology.

The problem is the "you only live once" (YOLO) approach that often characterizes the first stages of technology adoption. AI has been rolled out with a "move fast and break things" mindset, with no management and optimization frameworks to govern its use.

Larridin customers, who do have that framework, are able to rightsize their spending in line with productivity. We can offer a broad recommendation, and an encouraging result, from their experience:

• Employers need to set guardrails around the cost/benefit inflection point. With a management framework like Larridin, employers can quickly give employees useful feedback.
• Optimal productivity can be achieved at far less cost. With solid cost/benefit information in hand, employees can use their skills to get more work done for less.

We've used the same insights that power the report to launch new features under the banner, Token Spend & Insights. Click to learn more about these powerful new capabilities.

Want to know more? Contact us to receive a free copy of our report and for the opportunity to participate in a Larridin demo.