Summary of Key Points
Many large companies now use the amount of tokens consumed by employees when using AI tools (in simple terms, the number of AI calls or computational tasks) as a performance metric. Some companies provide free tokens but still rank employees based on their usage, while others require employees to subscribe to AI tools at their own expense. The reason behind this is that companies are unable to accurately measure the value that AI actually creates; they use tokens, which are easy to track, as a substitute. Employees are aware that it’s essentially a “numbers game,” but they feel compelled to participate because not doing so could lead to being perceived as “out of step with AI trends” and even affect their job security. As a result, everyone is caught in a new competitive race with no clear end in sight and no clear understanding of the benefits of winning.
Detailed Analysis
1. Why do companies use tokens as KPIs? – Because they are easy to track, even if inaccurate
Bosses advocate for embracing AI, but the biggest challenge is figuring out how much it actually helps with business goals. For example, whether a project was completed ahead of schedule is due to AI or because the requirements became simpler is difficult to determine precisely. Tokens, on the other hand, are straightforward: each AI call increases the token count, and running an AI model overnight shows an upward trend in the data. They provide clear evidence that employees are using AI. Although tokens don’t answer the question of whether AI has improved work efficiency, managers choose them because they save time—a performance metric that previously required reviews and project reflections can now be determined simply by exporting token data.
Here’s a relevant concept from management theory (Goodhart’s Law): Once a metric is used for important decisions, people will find ways to manipulate it. For instance, Amazon’s employees discovered that running useless tasks could boost their token rankings, leading to increased company costs, which prompted the company to abandon this metric.
2. Why do employees participate in the token competition? – Not participating may result in losing their jobs
Employees are not naive; they have calculated the costs:
- Low cost of compliance: They can use free company tokens multiple times or subscribe to AI tools at their own expense (which, although costly, is better than losing their job).
- High cost of non-compliance: Almost all companies now ask if employees know how to use AI, and this question persists even during job transfers. Being considered “out of step with AI” could put them at risk of being laid off.
Therefore, even though they realize it’s a pretense, they have no choice but to participate, as not doing so carries significant risks.
3. Why do companies allow employees to pay for AI tools? – Because it resembles personal consumption
Have you ever seen a company requiring employees to buy computers at their own cost? Probably not—computers are considered capital assets that the company must pay for. However, AI tools are different:
- They are subscribed to monthly, with additional charges based on token usage, similar to paying for video subscriptions or fitness memberships.
- Companies don’t explicitly state this, and employees are unsure whether the costs are their own or part of the company’s expenses.
This ambiguity allows companies to shift the financial burden to employees. For example, programmers in Shanghai may have to pay $2000 per month for AI tools, which is comparable to rent.
4. Are token rankings linked to layoffs? – Meta’s legal issues highlight the risks
When companies start laying off employees, token rankings stop being a “game.” For instance, Meta laid off 8,000 people this year, and 26 employees sued the company, claiming that token usage and activity levels were used to select targets for layoffs. Although Meta denied the claims and the court did not stop the layoffs, this incident shows that tokens can become a basis for termination if recorded in the system. Employees are forced to compete because they fear being labeled as “lagging behind” if they stop.
5. There’s no end to this new competitive race: Even if you win, what does it mean?
Employees jokingly refer to themselves as “chat engineers,” spending their days typing into AI interfaces. They worry that programming jobs may disappear by 2027, but they’re more concerned about the unknown benefits of winning token rankings—whether it means a raise or avoiding layoffs. Even Meta executives who advocate for unlimited use of AI didn’t make it into the top 250 in the token ranking. This indicates that there’s no definitive “right answer” in this race. No one dares to stop, as it’s like running in the dark, with wolves possibly chasing them.
In Conclusion
This token competition is essentially a way for companies to use easy-to-track numbers to avoid the complex issue of measuring AI’s value, shifting all the pressure onto employees. Everyone is participating, but no one knows what the ultimate outcome will be.
(End of article)