AI Token Budgets: The New Employee Benefit You Should Be Asking For

You’re reading an issue of "The AI Economy," my newsletter exploring the forces shaping the AI era—tracking how AI is rewriting business, work, technology, and culture. Subscribe to get expert insights and curated updates delivered straight to your inbox.

In Frank Herbert’s Dune, the axiom is simple: “He who controls the spice controls the universe.” In the AI era, the spice has a name—and it’s tokens. And as the technology continues to proliferate throughout our lives, these atomic units of AI are becoming as embedded in daily life as the dollar is.

Every AI interaction—every question asked, every answer generated—burns through tokens. It doesn’t matter if you’re asking it to help answer emails, summarize meeting transcripts, vibe-code an app, get a recipe suggestion from a smart assistant, translate a menu, summarize someone’s medical history, detect fraud, or resolve a customer support ticket. We are becoming a token-addled society, and we’re going to look for new ways to obtain more.

And when something becomes that essential, people start asking who’s paying for it. Take software engineers, for example, who are perhaps the most active category of AI users. Tokens are the lifeblood of their tools. Running out is like being handed a company car with no gas card—and being expected to fill the tank yourself. After a while, engineers may become fed up and look elsewhere.

Charles Lamanna, Microsoft’s executive vice president for agents and business apps, made the point bluntly at a GeekWire event* this week: “Imagine if you showed up to work and they were like, ‘Oh, you don’t get a mouse, just a keyboard…Oh, by the way, no emails and no [Microsoft] Teams. You just have to walk to people, or you can fax.’ You would go, ‘What the heck is this company?’ You would not stay around.”

He recounted a recent conversation he had with a potential job candidate who told Lamanna, “I will come as long as, in my team, they’re given at least X dollars of tokens per day.” While he didn’t disclose the exact amount requested by the applicant, Lamanna suggests it may have been upwards of hundreds of dollars per day and that it wouldn’t have been a bad move. “If an engineer is like $500,000…fully loaded, $100,000 of tokens a year to make them three times as efficient, that’s a great deal for everyone involved.”

While this situation may seem to be exclusive to technical professions, it probably won’t be for long. AI is infiltrating every role and aspect of work—entire organizations are being transformed. Compounding this is the rise of the agentic workforce. In an environment in which humans and bots collaborate, tokens aren’t just a personal resource—they’re now the payroll for digital labor.

And with AI being pushed onto workers, subjecting them to limited resources necessary to do their jobs can lead them to seek better opportunities. If they’re expected to use AI to improve their productivity and streamline workflows, how can it be done if companies are not giving them the tokens needed to succeed?

“How we think about what it means to hire, fully loaded costs, and where we invest, is going to change completely as a result of this,” Lamanna claims. “It’s happening in coding right now. It’s going to happen to every other form of office and information work over the next two years.”

He’s not the only one talking about tokens, either. NVIDIA’s Chief Executive, Jensen Huang, raised the issue earlier this month, proposing a new compensation model that would give engineers a token budget on top of their base salary. “I could totally imagine in the future, every single engineer in our company will need an annual token budget,” he remarked during his GTC keynote. “They’re going to make a few hundred thousand dollars a year—their base pay. I’m going to give them probably half of that on top of it as tokens, so they could be amplified 10x. Of course we would. It is now one of the recruiting tools in Silicon Valley. How many tokens come along with my job?”

However, token incentive programs may not be feasible for every business. It certainly appears more prevalent in technology firms such as Meta, Microsoft, Amazon, and Google. It’s becoming common to hear executives from Big Tech talk about how their developers are no longer actually coding. Rather, the work is being done by OpenAI’s Codex, Anthropic’s Claude Code, AWS’s Kiro, Microsoft’s GitHub Copilot, or some other coding agent.

“For large portions of my team, I have products which have incredible scale and massive revenue, where a human does not write a single line of code directly anymore—zero lines of code written by a person directly in the code base,” Lamanna claims.

That’s all well and good, but tokens aren’t free. And for companies adopting AI, it can be a huge bill to swallow. Box Chief Executive Aaron Levie urges business leaders to start discussing how to budget for these AI token bills, especially as usage begins to skyrocket across the entire workforce.

The irony is that the same technology companies that are pushing onto their workers are the ones whose technology those workers will increasingly demand be funded for them. Token budgets are both the incentive and the overhead, but who will pay for them? The question businesses need to answer soon is whether they’d rather budget for it proactively or lose their best people to someone who already did.

Featured Image: Charles Lamanna, executive vice president of agents and business applications at Microsoft, speaks onstage with Todd Bishop of GeekWire during the “Agents of Transformation” event in Seattle on March 24, 2026. Credit: Ken Yeung