Hello! 👋
It's Thursday, 18th June 2026. Hello and welcome back to Bold Efforts!
Some companies are starting to track which employees use the most AI. In a few cases, this turns into an internal leaderboard. The people at the top look like early adopters. The people at the bottom get nudged to catch up.
I find this more revealing than the companies running it probably intended.
A token is a small unit of text, often part of a word. When you ask an AI model to write, analyze, summarize, code, or respond, its work is measured in tokens. Tokens cost money, but for most serious companies, the cost is not the real issue.
The real issue is what the leaderboard measures.
A token leaderboard measures who generated the most text. That is not the same as who did better work. You can produce a million words of AI output and still make no sharper decisions, write no clearer strategy, close no better deals, and ship nothing that would not have existed without the AI.
More usage is not the same as more leverage.
The anxiety behind this trend is real. Boards are asking executives whether their companies are AI-ready. Investors are asking the same question. Every major technology company is reorganizing around AI. Every leadership team knows it cannot be seen as slow.
So executives need something to show.
The token leaderboard becomes that answer. It is easy to collect. It is easy to report. It creates the appearance of transformation without forcing anyone to ask the harder question: is the work actually getting better?
This is older than AI.
For decades, companies used presence as a proxy for value. Who arrived early. Who stayed late. Who was visible to the people who made decisions. Remote work disrupted that habit, and for a short moment, companies had to evaluate output instead of attendance.
Many could not.
So they found new inputs to measure. Camera-on time. Active hours on Slack. Response speed. Calendar density. The format changed, but the failure stayed the same.
Token leaderboards are the AI version of staying late. They prove you were in the building. They say nothing about what you built there.
The companies learning to use AI well are asking a different question. They are not asking, “Who used the most AI?” They are asking, “What work should AI absorb, and what should humans do better with the time it frees?”
And questions like these are more important.
If AI is working, sales teams should spend less time drafting follow-ups and more time understanding customers. Analysts should spend less time formatting memos and more time improving judgment. Managers should spend less time summarizing meetings and more time making decisions. Product teams should spend less time writing status updates and more time removing ambiguity.
You can measure those changes.
Did the sales cycle shorten? Did support teams resolve issues faster? Did strategy documents become clearer? Did engineers ship better work with less coordination drag? Did leaders make better decisions with fewer meetings?
That is AI adoption. Token count is only activity. But we often mistake motion for progress.
The absurd part is that most organizations already have too much text. More emails than decisions. More summaries than insight. More documentation than understanding. AI’s immediate value is not that it helps us produce infinite text. Its value is that it can compress the unavoidable text and return attention to the work that needs judgment.
A token leaderboard pushes people in the opposite direction.
It tells them to generate more. More prompts. More drafts. More summaries. More visible usage. It turns AI into another performance ritual, another way to prove you are participating in the company’s current obsession.
And once that happens, people learn the game.
They will run longer prompts. They will generate more drafts than they need. They will ask AI to summarize things they already understand. They will optimize for the metric because that is what people do inside systems that reward metrics.
The leaderboard will go up. The work may not.
The deeper problem is not tokens. The deeper problem is that many organizations have forgotten how to recognize good work without a number attached to it (hours in office, hours online, number of lines of code etc.)
If you can evaluate whether AI is improving your team’s output, you do not need to count tokens. You look at the output. You look at the decisions, the speed, the clarity, the quality, the customer impact.
Token counting is what you do when you cannot evaluate the work. And you cannot evaluate the work when you no longer know what good looks like. That is the real confession inside the leaderboard.
The companies that figure out AI will not know it worked because usage went up. They will know it worked because the work changed. Less noise. Better judgment. Faster decisions. Clearer output. More time spent on what only humans can do.
When you cannot evaluate judgment, you count actions. That has always been true. AI just gave the problem a new unit of measurement.
Thank you for reading!
Best,
Kartik
I write Bold Efforts every week to think clearly about where work and life are actually headed. If you want these essays in your inbox, you can subscribe here.