I remember the night AI’s impact at work clicked for me.
I was working during Double 11 at Alibaba, China’s version of Black Friday but roughly ten times the scale. The numbers are almost impossible to describe: hundreds of millions of shoppers, billions of dollars transacting in a single day, and a creative team responsible for producing the visual assets to support all of it. Banners, product images, promotional graphics, personalized storefronts. Thousands of them. In hours.
What I watched that night was a generative AI design tool do in seconds what used to take a team of designers days. The tool was generating thousands of banner variations, adapting layouts, colors, and copy for different customer segments in real time.Â
That was the moment I knew that the future of work w0uld look a lot different from what we know today.Â
At the time, what didn’t click for me was that tokens would become the new form of labor, because at the time that was an internal tool in beta. But fast forward to today, where AI systems are such common work tools, things are starting to look a lot different. Â
The Salary-Token TradeoffÂ
At NVIDIA’s GTC keynote in March 2026, CEO Jensen Huang floated a concept that would have sounded like science fiction five years ago: offering engineers half their salary in the form of AI compute tokens.Â
According to CNBC’s coverage of the event, Huang framed tokens as an attractive hiring mechanism, a way for employees to multiply their output well beyond what a paycheck alone could buy.
No major company has publicly implemented this yet. But the idea is worth sitting with. Compensation is evolving beyond money for time. Increasingly, it includes compute power, and what matters for career advancement is what you can produce and at what scale.
Picture a marketing professional weighing two job offers.Â
- Company A: $150,000 plus one billion AI tokens.Â
- Company B: $180,000, no tokens.Â
At first glance, Company B looks better. More money, full stop. But a billion tokens represent real leverage. With that compute allocated to you, you are functioning less like a solo marketing manager and more like a one-person department.
Don’t fall into the other trap either. Before you say you’re great with AI and accept the lower paying job, a smart candidate should be asking deeper questions. What model does the token package give access to? Can you run autonomous agents? Can you build internal tools? Can you deploy those tokens on side projects? Are there restrictions on what you produce?
Labor is starting to look like an AWS billÂ
To understand why this tradeoff matters beyond individual career choices, it helps to look at how companies currently account for labor.
When a company hires a full-time employee earning $120,000 with $30,000 in benefits and taxes, it commits to roughly $150,000 annually, regardless of output. Whether that employee processes 20 customer tickets or 2,000 in a year, the cost stays the same. The real question becomes utilization: what is the cost per task completed?
For a support agent handling 20,000 tickets annually at a total cost of $150,000, the effective rate works out to $7.50 per ticket.
AI operates on a fundamentally different model. Instead of paying for time, companies pay for usage. Costs are driven by tokens processed, compute consumed, and API calls made. Each task has a direct, measurable cost associated with it.
For instance, a typical AI-powered customer support interaction might involve 1,000 tokens to process the input and 500 tokens to generate a response, for a total of 1,500 tokens. If the model costs $0.005 per 1,000 tokens, the total cost of that interaction is approximately $0.0075, less than a cent. Compared to the $7.50 cost of a human-handled ticket, this represents a difference of several orders of magnitude.
CFOs are doing this math. The conclusion is that labor, historically a fixed cost, is becoming a variable one, closer to a cloud computing bill than a payroll ledger.Â
Companies are beginning to operate on what might be called AWS billing logic: pay for what you use, scale what produces results.
This creates pressure not just on headcount, but on how headcount is valued.
Who Gets Left Behind
This is where the long-term consequences become serious, and where the future of work conversation needs to go beyond efficiency gains.
I use a simple framework to think about how work divides in an AI-powered organization: bees and beekeepers. The bees are AI, executing tasks at scale, processing inputs, generating outputs, operating continuously without fatigue. The beekeeper is the human, directing the hive, setting priorities, making judgment calls, and deciding what the work is actually for.
The workers most at risk in the token economy are those whose entire job description has, until now, looked like bee work. Data entry, routine reporting, basic customer support, document processing. These roles require execution, consistency, and volume. Historically, humans filled them because humans were the only option. With AI handling the same tasks at a fraction of a cent per interaction, that rationale is gone.
Those agencies Alibaba hired every Double 11 were full of people doing bee work. Producing at volume, to spec, on deadline. The gen AI tool that replaced them did not need a budget line, a contract, or a project manager. It needed a few clicks.
These roles also matter beyond economics. They have historically offered a point of entry for workers without advanced degrees, workers building skills over time, workers in geographies where white-collar wages are lower but still meaningful. If the cost of automating them drops to fractions of a cent per task, the economic case for keeping humans in them erodes quickly. The workers in those positions are not the ones who will be offered token packages as a retention incentive. They will simply find fewer positions available.
Become the BeekeeperÂ
The workers positioned to benefit from AI-powered compensation models are those who were already operating at the beekeeper level: people who direct, evaluate, build, and decide. For them, a token allocation multiplies what they can already do. For workers whose value has been tied entirely to execution, the arrival of AI does not offer a multiplier. It offers a replacement.
This is the harder conversation most organizations are avoiding. Retraining people from bee tasks to beekeeper responsibilities requires time, investment, and a genuine commitment to workforce development. So far, the urgency around cost savings has moved considerably faster than the urgency around that transition.
The Question Worth Asking
Would you take less salary for more AI power? Before you answer, ask yourself a harder question first: are you the kind of worker who would know what to do with it?
The token economy rewards beekeepers. People who direct, evaluate, decide, and build. People who can look at a billion tokens and see leverage rather than a perk. For workers who have spent their careers executing bee tasks, a token package is not an opportunity. The conversation about compensation structures will simply pass them by, while the conversation about their role’s continued existence happens somewhere else entirely.
Future-proofing against AI disruption is not about learning to use more tools. It is about moving up the hive. The workers who come out ahead will be those who stop asking “how do I do this task?” and start asking “how do I decide which tasks get done, by whom, at what cost, toward what end?” That is the beekeeper question. And right now, not enough workers are asking it.
