By Ingrid Schneider, Founder — Train in Your Lane
I was raised in a mall.
That’s not a metaphor. My dad owned waterbed stores, so my childhood happened in food courts and back hallways and, most importantly, the arcade.
The arcade was glorious. My dad would hand us one five dollar bill. One bill. Three kids. If you want to know why I’m good at negotiation, start there.
We’d feed that five into the change machine and out came tokens. Real brass tokens, only good at that arcade, only good for those games. And every mall kid learned the same brutal economics: when the tokens ran out, the games stopped. Didn’t matter that you were three ghosts deep into your best Ms. Pac-Man run. Didn’t matter that the barrel was in the air and Donkey Kong was RIGHT THERE. No tokens, no play.
You went back to the change machine or you went home.
Congratulations. You now understand enterprise AI pricing better than most executives I meet.
The $25 isn’t the product
When your company pays Claude $25 a month per team member, you’re not buying “AI.” You’re buying tokens. A big pile of them, but a pile.
Every prompt costs tokens. Every response costs tokens. Upload a 40-page PDF and ask for a summary? Tokens. That marathon conversation where the AI has to re-read everything you’ve said just to answer your next question? You don’t want to know.
And when you hit your limit, the machine does exactly what the arcade did. It stops. Mid-game, mid-project, mid-brilliant-idea. Come back later or buy more.
The subscription is the five dollar bill. The tokens are what you’re actually spending. Nobody explained that to your team, which is why half of them are playing Donkey Kong with the company card.
Watch what just happened at Anthropic
Here’s why the language matters.
Since May, Anthropic has been running Claude Code weekly limits 50% higher than standard. A promotion. Extended three times, most recently through August 31.
Then in late August they announced the promo ends September 14 — and standard limits go up 25%. Permanently.
Read that fast and it sounds like a raise. Permanent increase! More tokens! Free play!
Do the math and it’s a cut. You’ve been living at 150%. The new normal is 125%. That’s 17% of your capacity gone — capacity your team has spent four months building workflows on top of. Anthropic confirmed the figure themselves once people ran the numbers: “Compared to today, this works out to a 17% reduction in weekly limits on Claude Code.”
Anthropic didn’t lie. They announced an increase, and technically it is one — measured against a baseline nobody has actually used since May. That’s not fraud. That’s fine print. The arcade never lied to us either. The token machine just sat there next to a sign that said 25 TOKENS FOR $5 and let us believe that was a lot.
If you don’t understand tokens, promotions, baselines, and limits, you can read an announcement that says “more” and end up with less. That’s not an AI skill. That’s a reading-the-actual-numbers skill. Your CFO would call it Tuesday.
People are using AI like Google
And that’s the real problem.
Most teams treat AI like a search bar with better manners. Type a question, get an answer, close the tab.
I have the receipts. Before a recent executive training, we surveyed the room — 22 leaders, most using AI several times a day. Confident, too: 18 of 22 rated their own AI skills a 4 or 5 out of 5.
Then we asked one open question: when you type something into ChatGPT and it answers, what’s actually happening?
14 of 22 said some version of “it’s searching the internet.”
It’s not. It’s not searching anything. These models generate language token by token based on patterns in training data — which explains the rest of the survey, because 12 of the 22 couldn’t tell you what training data is. Seven believed AI does math reliably. Five believed it remembers every past conversation perfectly.
Eight out of ten confidence. Four out of ten understanding.
That gap is expensive. If you think the tool is Google, you’ll prompt it like Google, trust it like Google, and burn your entire token budget getting Google-quality answers from a machine that could have been drafting your franchise brochure.
That’s not using AI. That’s putting a token in the machine and only pressing the start button.
Teach the team, not the tool
This is why we built organizational training at Train in Your Lane. Not “here’s where the button is” training. Training on how these systems actually work — what a token is, what a model does, why the confident answer is sometimes completely made up — so your team stops button-mashing and starts playing like they’ve seen the pattern.
Because here’s what every arcade kid knew: the high scores didn’t go to the kids with the most tokens. They went to the kids who understood the machine. The kid who knew the ghosts in Ms. Pac-Man have patterns. The kid who knew when the barrels come down the ladder.
My dad gave us five dollars. It went a lot further once we stopped playing games we didn’t understand.
If your team is using AI every day and nobody can explain what a token is, that’s the gap we close. Train in Your Lane runs live AI training for organizations — your team, your tools, your actual workflows. One session and your people stop guessing. Reach out and let’s put a date on the calendar.
Most companies are still standing at the change machine wondering where the money went.
Don’t be most companies.
book a discovery call — free · 30 minutes
