- AI Business Insights
- Posts
- The 120 dollar split
The 120 dollar split
Forty comparison posts, zero decisions
In partnership with
You know the genre. Two logos, a table of benchmark scores, a paragraph about how it really depends on your use case, and then nothing. I have burned entire evenings on those posts and come away with exactly what I walked in with, which is a browser tab I do not want to look at anymore.
So when I found a breakdown from somebody who had stopped comparing and just paid for both, ran them side by side on actual work for a stretch, and then wrote down what he learned, I read the whole thing twice.
His conclusion is one sentence:
Pay Claude $100 a month, it does the work. Pay ChatGPT $20 a month, it makes the images.
A hundred and twenty dollars, and he says he would spend it exactly the same way again. What follows is his reasoning, and the part that surprised me is that almost none of it is about model quality.
How Jennifer Aniston’s LolaVie brand grew sales 40% with CTV ads
For its first CTV campaign, Jennifer Aniston’s DTC haircare brand LolaVie had a few non-negotiables. The campaign had to be simple. It had to demonstrate measurable impact. And it had to be full-funnel.
LolaVie used Roku Ads Manager to test and optimize creatives — reaching millions of potential customers at all stages of their purchase journeys. Roku Ads Manager helped the brand convey LolaVie’s playful voice while helping drive omnichannel sales across both ecommerce and retail touchpoints.
The campaign included an Action Ad overlay that let viewers shop directly from their TVs by clicking OK on their Roku remote. This guided them to the website to buy LolaVie products.
Discover how Roku Ads Manager helped LolaVie drive big sales and customer growth with self-serve TV ads.
The DTC beauty category is crowded. To break through, Jennifer Aniston’s brand LolaVie, worked with Roku Ads Manager to easily set up, test, and optimize CTV ad creatives. The campaign helped drive a big lift in sales and customer growth, helping LolaVie break through in the crowded beauty category.
*Ad
Move your setup first, then test
This is the tip I wish somebody had handed me months ago, because I have done the dumb version of this at least three times.
When you try a new tool, the instinct is to start clean. New account, blank slate, let me set this up properly. So you spend a week rebuilding your instructions, re-pasting your brand voice, reteaching it your formats. By the time it works you have forgotten what you were even comparing.
He does the opposite. He takes an existing Claude skill and hands it over with one line:
Make a skill from this [Skill name]. Just update Claude to ChatGPT.
That is the whole migration. Your configuration transfers, and you start at expert level instead of fumbling through beginner mode for a week.
There is a second thing happening here that I think matters more than the time saved. If you rebuild from scratch on the new tool, you are not comparing two tools. You are comparing a setup you have tuned for months against a blank account. Of course the old one wins. Port the setup and the test is finally fair.
Here is where most people quietly give up, and honestly the arithmetic explains why.
He counted the options: 2 models, times 3 flavors, times a range of effort levels from Light to Extra High, times 2 speed settings. Around thirty combinations. That is not a product decision, that is a diner menu.
His filters cut it down fast.
Fast mode costs 1.5x. Not a rule, just a question to ask on each task. Is the speed worth the premium right here, for this specific thing? Usually not.
Skip the middle flavor entirely. His reasoning made me laugh out loud, because nobody actually knows what a medium task is. Middle options exist so the lineup looks complete, not because anyone picks them on purpose.
Plan on the heavyweight, execute on the workhorse. This is the real find. The top-tier model at max effort devours your usage. So he runs a couple of turns on the expensive one to plan the work, then drops down to the efficient model to actually do it.
That last one is a genuinely different way to think about model selection. You are not picking a model for the project, you are picking a model per phase. Thinking is a small number of expensive turns. Execution is a large number of cheap ones. Match the spend to the shape of the work and your usage limits stop screaming at you by Wednesday.
The best voice models, now fully orchestrated across all channels
ElevenAgents puts full orchestration on the voice models the market builds around. Voice, text chat, transcription, and reasoning in one integrated stack, <400 milliseconds, and human-sounding. Plug in any LLM, integrate tools, A/B test, and deploy across channels. More human-like conversations, lower latency, flat $0.08 per minute.
*Ad
The 151st hire
This section is the one I would print out and hand to a finance person, because most teams find these numbers only after they have signed something.
ChatGPT Enterprise has a 150 seat minimum. So a 150 person company is looking at roughly $3,000 a month in seats plus around $17,000 a month in tokens. Run the heavyweight model on high effort across that team and he puts the real bill near $37,000 a month.
Claude under 150 seats: $100 flat.
The catch is seat number 151, which tips you straight into pay-per-token territory.
His two rules:
Under 150 people, take the flat seats now, before the pricing moves.
Anywhere near 151, budget for tokens before you hire, not after.
Sit with that for a second. Your AI bill can jump twelve times on a single new hire. That is not a pricing tier, that is a cliff with no railing, and it sits at a headcount number that a growing company crosses without ceremony. Somebody signs an offer letter on a Tuesday and the software budget changes shape.
I do not think most teams have any idea where their cliff is. Worth twenty minutes to find out where yours sits.
One account for the whole team
Small, cheap, and it works because it matches how the tool actually gets used.
He does not buy an individual seat for everyone who occasionally needs an image. One shared $20 account covers the team.
Think about how image generation shows up at a normal company. Someone needs a header for a deck on Thursday. Someone else wants a rough mockup before a call. It is bursty, not constant. Paying per person for something that sits idle five days out of seven is how software budgets balloon without anyone deciding to spend more.
The general version: match the license model to the usage pattern. Constant daily use, buy seats. Occasional bursts, share one account. That question is worth asking of half the tools on your invoice, not just this one.
37 Free Claude Prompts With The AI Report
Subscribe to The AI Report, the free 5-minute daily AI brief for 400,000+ business leaders, and you’ll get 37 Claude prompts free in your welcome email. They’re organised by the 8 situations every manager faces. You get both: the newsletter and the prompts.
*Ad
When you genuinely cannot decide
My favorite part, because it cuts directly against everything the comparison industry is selling.
His tiebreaker: pick the one your team already opens every day.
He makes the point that every company has one AI superman. One person who is already deep in a tool, already building things with it, already dragging colleagues along in their wake. Follow that person. In his words, adoption beats every benchmark.
I think that is exactly right, and it is the kind of thing that sounds soft until you watch it play out. A slightly weaker model that people actually open will outproduce a marginally better one sitting in an unclicked bookmark. Tools do not produce value at the point of purchase. They produce it at the point of habit, and habit already has a winner in your company whether you have noticed or not.
The sequence, if you want to run this yourself
Port your existing skills and setups to the second tool with his prompt, so both sides start even.
Ignore the middle-tier options. Pick one heavyweight and one efficient workhorse.
Plan on the expensive model, execute on the cheap one.
Check your headcount against that 150 seat line before you sign anything.
Share one low-tier account for bursty work like image generation.
When the analysis stalls out, go with whatever your team already opens.
What I actually took from this
The thing that stuck with me is not the $120 or any of the specific numbers, because those will be wrong within a year. It is that he answered the question by running the tools instead of reading about them.
And notice what his answer is actually made of. Seat minimums. Usage patterns. Which model to use for which phase. Who on the team already has the habit. Almost none of it is about which model is smarter. As pricing splits into flat seats versus consumption billing, the useful question stops being which model is best and becomes which billing model matches how my team actually works.
That is a less exciting question. It is also the one with money attached.
Tonight, do the smallest piece: open your last invoice and count your seats against 150. Five minutes, and you will either sleep fine or you will have found something worth knowing before your next hire does it for you.