It’s the buzzword of the moment. Nobody doubts anymore that running out of tokens is the equivalent of running out of credit, missing the bus, or cancelling your holiday. But what exactly are tokens, and why is everyone talking about them? In this article, beyond trying to explain tokens in a way that anyone can understand, I want to share with you some paradoxes I see every day in Spanish companies. But I also want to show you how countries like the United States are ahead of the curve — where the word “token” is already being used in lifts, taxis… and of course in job interviews. Keep reading, there’s much more to come.
Not so long ago, when a candidate was negotiating their joining package, the benefits on the table were the usual suspects: meal vouchers, a company car, private health insurance, maybe a few executive coaching sessions if the company was one of the good ones… Well, in many tech companies in the United States, a new line has appeared in those negotiations that just two years ago would have sounded like science fiction: tokens. Unlimited tokens, or at the very least, a generous monthly package of access to large language models. Not without my tokens. Plain and simple.
So what exactly is a token? Put simply, a token is the unit of measurement that artificial intelligence models like Claude, GPT-4 or Gemini use to process language. Roughly speaking, three tokens are equivalent to one word in English. When you interact with an LLM (like Claude, which is my personal favourite right now), every question you ask and every response you receive consumes tokens. And tokens, depending on the model and the volume, cost money. That’s why in companies with a high technological workload, the question is no longer just how many people do we need, but how many tokens do we need for those people to work effectively.
Tokens have become the new measure of productivity. And that has changed everything in the human resources equation.
The silent transformation of the workplace
AI has gone from being an optional accelerator to becoming the main engine of every professional function. Analysts, lawyers, doctors, engineers, salespeople, marketing directors: all of them have seen how tasks that used to take hours of work are now resolved in minutes when you have the right tools — and the tokens to use them.
Just this morning my physiotherapist was telling me that at the last conference she attended at the beginning of the year, everything was about technology. There wasn’t a single scientific talk. It was all AI, tokens, LLMs, robotics… Are we living through the end of the world as we know it today?
In the United States, the most advanced companies no longer hire only human heads (or as the cool kids say, headcounts). They also hire digital heads: AI agents that work 24 hours a day, 7 days a week, 365 days a year. No sick leave, no conflicts with works councils or trade unions, no endless annual pay reviews, no need for health insurance. As I already wrote in my article on Agentic AI, the most advanced HR departments no longer talk only about FTEs (Full Time Equivalents) — they also carefully consider how many digital agents accompany each real person. And this, to be honest, comes from someone who thinks the most complex part of leading a team is the people. Not the worst part: the most complex. For the record.
The Spanish paradox: more AI, less productivity
And here comes what concerns me most, which I already explored in my article on the silent paradox of AI. In Spain, many companies are incorporating artificial intelligence tools in a tactical rather than strategic way. They are not redesigning processes or restructuring their org charts. They are applying patches.
And what happens? Instead of eliminating work, they are duplicating it. Reports are being generated that didn’t exist before. And those reports need to be reviewed. And based on those reviews, decisions need to be made that didn’t exist before either. The result is paradoxical and absolutely real: greater individual speed, greater collective slowness. We are installing Formula 1 engines in structures designed to travel at 60 km/h.
Hiring decisions in many Spanish companies are still being made using the same criteria as always, completely ignoring the fact that the token variable — that is, real and sufficient access to AI tools — is today just as relevant as salary. A professional with limited access to LLMs in 2026 is like a professional without a computer in 1995. They function. But much worse.
The new roles that are already arriving
In the United States, profiles are emerging that in Spain still sound like a distant future: AI leads, output reviewers, strategic decision-makers working from AI-generated information. People whose value lies not in executing tasks, but in supervising, validating and deciding on what agents produce. At the same time, purely administrative roles — those whose added value was limited to processing information — are disappearing. Not because AI is magic, but because it does that work better, faster and without making the same mistakes.
The question that HR directors — or People directors, as I prefer to call them — need to be asking themselves today is not how many people do I need. It is: how many tokens do my people need to perform their role effectively? And how many digital agents can work alongside them?
Whoever understands this first will win. Whoever keeps negotiating only company cars and meal vouchers will arrive too late.
As allways
Are you already negotiating your tokens in your recruitment processes? I’d love to read your thoughts in the comments.