The frontier gap tripled in six months

It’s almost Friday, hang in there!

The gap between companies that are good at this and those that bought licenses has tripled since January. OpenAI published two studies together on 12 August. Frontier firms, the top 10% by usage in a given month, generate 8.3 times as many output tokens per active user as typical firms. In January, that gap was 2.6 times.

Tokens are a crude proxy for depth of use, and OpenAI is measuring its own customers against its own metric. Say that out loud before the number goes in a deck.

What sits underneath is less crude. Among weekly active users, 21% at frontier firms use Plugins and 19% use skills, against 9% and 3% at typical firms. At OpenAI itself, 95% use Plugins weekly. That is the difference between typing into a chat box and agents wired to company data, tools, and a repeatable workflow.

Then the finding your leadership team will not enjoy. Usage is highest among early-career employees and falls as seniority rises. The people writing the AI strategy are using it least.

Marketing is not the laggard for once. Weekly active enterprise Codexusers are up 26 times in marketing since February, against 5 times inengineering.

Marcio Arnecke, CMO of Apollo, on his own survey:

He is right about the hurdle. In the same survey, 84% use AI for prospecting and research, the easiest thing in the funnel to hand over. Adoption stopped being the interesting question a year ago, and most teams have not replaced it with a better one.

Near-universal adoption, one in six measuring it. The gap between those two numbers is the job.

Use this when your license count is high, and your output hasn't moved. The seats were the cheap part, and they were never the constraint.

Kick is self-driving bookkeeping — transactions get categorized in real time, the home office and travel deductions stop slipping through, and the P&L sits tax-ready instead of being reconstructed in a panic every April.

It runs across unlimited entities, so the consulting arm, the side product, and the rental all reconcile in one place, with rules you can shape to how your money actually moves.

New users get 10% off for 12 months through our link.

The frontier gap is not a talent gap. It is the difference between forty people who each have a good private workflow, and a team where one of those workflows became the way everybody works. Nobody will ever put the second thing in your objectives, and it is the only part of this that compounds.

We provide CMOs with AI strategy consulting for marketing teams through our Teamless offering.

We run an interview series with marketing leaders about how AI actually changes their work. Real use cases, not predictions.

If you’re doing something worth talking about, the opportunity is free and available for anyone. Simply reply to this email with your pitch.

The frontier firms are winning on plumbing. Three ways to lay some without a headcount request:

Gumloop — best for wiring your existing marketing tools to agentswhen nobody on the team writes code. Try Gumloop here →

Stilla — best for one permission-aware teammate working across thecompany rather than a separate bot per tool. Try Stilla here →

Viktor — best for the recurring recap that lands in Slack with thenumbers pulled, before anyone asks. Try Viktor here →

We have 300+ curated, filtered, and categorized AI tools in our directory. Browse to find the best ones that help your marketing team.

We negotiated discounts, bonuses, or extended trials on AI marketing tools.

For $10/month, you save $1,724/month on tools. Plus, you get access to our full content archive & chat as a bonus.

We even offer a 10% OFF bonus on all paid plans for free subscribers.

If the frontier gap is an operating-model gap, start by being honest about where you sit: the AI Marketing Maturity Model Framework.

A structured assessment plus the roadmap out, built to move a marketing organization from experimentation to systems that show up in revenue. It earns its place by forcing you to write your current stage down, which is the argument you have been having sideways all year.

We publish free workshops, courses, guides, manuals, and more on AI marketing. Browse our learning resources for free.

Cliff Tan makes interior design videos for 2.4 million YouTube subscribers as Dear Modern. He ran a paid ad for Dreamina, ByteDance's AI creative tool, and tried to inoculate himself inside the ad, saying he does not "like AI" because "it makes you forget to use your brain."

It made no difference. The loudest objection in the comments collected 1,700 likes: "No, Mr. Cliff. Please. Don't sell yourself to AI." He published an apology video on Monday. "Yesterday I made a video on AI, and it disappointed many of you," he said. "I'm really sorry about that."

Second one that week. Hank Green, 3.2 million subscribers, apologized on 4 August for using ChatGPT to find papers and sources, then published a personal AI policy ruling out letting a model write, edit, or outline anything.

Neither of them generated anything wholesale. Both were punished for standing next to it.

So if you are buying creator distribution for an AI product, look at what you are asking someone to spend. The rate card prices their audience. It does not price their standing with that audience, and the creator settles that bill before you do.

That’s it for today!

See you next time,

— Peter & Torsten

Thanks for reading AI-Ready CMO! This post is public, so feel free to share it.