In partnership with
Help make better ads
Did you recently see an ad for Roku Ads Manager in a newsletter? We’re running a short brand lift survey to understand what’s actually breaking through (and what’s not).
It takes about 20 seconds, the questions are super easy, and your feedback directly helps us improve how we show up in the newsletters you read and love.
If you’ve got a few moments, we’d really appreciate your insight.
Hello again my friend,
I couldn’t believe it when I saw it:
A lot to unpack there. But it’s not the only topic for today’s Liight Work Weekly.
AI bills are skyrocketing and companies are pulling back.
The difference between how large companies use AI vs the rest of us.
Which companies are building AI for, again, the rest of us.
Which leads to a question on content, who do we trust (especially in marketing)?
If you are looking to learn more about how all this AI stuff works generally, (whether you want to start a business with it, or just try it), this post will bring you up to speed.
Let’s lock in.
The AI bill came due
There’s been a wave of posts about companies pulling back on AI tools because the bills are getting ridiculous. It’s sick to call it funny because people really lost their jobs over this, but you can’t ignore the irony.
They spent half a billion dollars in a month because they forgot to set their spending limits. I rarely cuss in the newsletter but that’s fucking wild. Just imagine waking up to that bill.
It’s not just this case either. Uber’s in the same boat where each developer was capped at $1,500. Even Sam Altman (CEO at OpenAI), is warning people, mostly investors, that it’s a thing.
But, how can it be a thing? And is everyone going to experience this soon?
Not quite.
Big companies vs “the rest of us”
If you use ChatGPT, Claude, Notion AI, Cursor, or any of these tools as an individual, you usually pay a fixed monthly fee. Maybe it’s $20/month. Maybe it’s $200/month. Either way, you know the number before the month starts. When your usage hits the maximum, you have to wait.
“Usage” has a few parts to it.
The time and effort AI spends on your tasks take up tokens. More time and effort means more tokens (this is your usage).
Plans come with usage limits. So for $20 a month, there’s a cap in the system already that’s respected so you’re not charged extra.
The limits for usage are set per session (a 5-hour block), per week, and per month.
So you can use up all your tokens in a session, but still have usage left over that week, you just have to wait for the next session to start. I imagine they do this to put throttles on everything (including their own data centers) so nothing overloads.
That’s typically for “the rest of us,” who do not have millions of dollars. Buy a $20 plan, use it, and then when tokens are used up, wait for a reset.
Massive companies do not have plans like us. They just go. Every employee (in the case above) had access to unlimited tokens, used AI for everything (probably not very well), and suddenly there’s a bill for half a billion.
So if 100 employees barely use it, the bill stays manageable. But if 5,000 employees start using it all day, every day, across coding, research, summaries, internal tools, customer support, data analysis, and whatever else, the bill can ramp quickly.
There are easy ways to set limits, but it looks like they just missed that part. Which brings up the next point. Is this a real issue, or did a few companies just mess up?
Those companies just messed up
A lot of people will read this and say, “See, AI isn’t enough.” or “I told you so, it’s not as good as it sounded…hahah…”
I hate to say it, but that’s just cope. It’s not legit and I would not take that kind of commentary seriously.
If anything, it’s probably the opposite. A few senior execs at these companies, who are probably near retirement and so far removed from the game, didn’t plan and manage all this properly and so they got a huge bill.
Will AI become more expensive one day? Definitely could, but it’s not today. Right now is actually the cheapest it’ll ever be because they’re heavily subsidized (Open AI loses money), it’s just that you have to manage it sometimes, and they didn’t. That’s why the issues above are so ironic.
Where to place attention:
In business:
This post from Zeb Evans is a great example of how successful CEOs are seeing AI. It sucks that he fired 22% of his staff, but the message underneath is important (and a smart marketing tactic on his part).
Recap:
The people who manage the AI and systems are becoming more valuable.
The ‘human touch’ is vital, do not replace it.
AI allows one person to make a 100x impact, and Zeb claims his compensation is going up to $1M for individual employees (sometimes higher).
The part no one says out loud is how ClickUp (their product) is 2 years away from people just making their own, specialized version of it with AI for less money. Don’t get me wrong, great product, but the world’s just going in a direction where this type of software is at risk.
In marketing:
This video from the Grant Owen podcast with Oren John is a great example of great content. More importantly, a great example of the kind of content creator that will win 100x more than the average.
Why do I feel so strongly about this?
Oren actually has nearly 1M followers on social media himself.
He’s been doing this for decades. He’s not a kid who woke up and said “I’m an expert,” and started selling courses.
What he’s saying works. I’ve tried frameworks from his other videos before.
So proof, experience, and credible information all come together to make it work. I thought it was good enough to share here for sure, so I hope it’s valuable to you.
If you’re considering starting a business by the way, it’s a good watch/listen.
In beehiiv
Our Summer Release event is July 16th > RSVP here.
We’re hiring a Senior Product Manager ($140k - $160k USD).
In case you missed it, we hit $30M in subscription revenue.
But that’s all for this week. As always, thanks for reading!