The 6x Divide

Why your coworkers are in a different AI economy than you and how you can catch up

Author

Justin Wright
December 22, 2025 • Est. Reading Time: 9 minutes

Happy Monday!

OpenAI recently dropped its first-ever enterprise AI report, and buried in the data is something nobody's talking about: workers at the 95th percentile of AI adoption send 6x more messages than the median employee at the same companies.

For coding tasks? 17x. For data analysis? 16x.

Same subscription. Same tools. Same training sessions. Completely different realities.

This isn't about access anymore. ChatGPT Enterprise is deployed across 7 million workplace seats, a number up 9x from last year. The tools are identical for everyone. And yet usage varies by orders of magnitude within the same organization.

We're not in an AI adoption story anymore. We're in a workforce bifurcation story.

OpenAI's enterprise report reveals a stunning productivity gap: frontier workers (95th percentile) use AI 6x more than median employees at the same companies (17x more for coding). Average workers save 40-60 minutes daily; power users save 10+ hours weekly. Meanwhile, 19% of active users have never tried data analysis, and 14% have never used reasoning models. The divide isn't who has AI, it's who's actually using it. And the gap is accelerating.

TL;DR

The Numbers That Should Terrify You

Let's establish what's happening at the top of the adoption curve.

ChatGPT now serves 800 million weekly users. Enterprise message volume grew 8x year-over-year. Reasoning token consumption per organization increased 320x. Organizations aren't just using AI more; they're using it for increasingly complex, multi-step tasks.

The average enterprise worker saves 40-60 minutes per day. Data scientists and engineers report 60-80 minutes. And for heavy users? More than 10 hours per week. That's essentially a free extra workday.

And 75% of workers say they're completing tasks they literally couldn't do before like coding, data analysis, and custom workflows. Coding-related messages from non-technical workers jumped 36% in six months. Marketing people are writing scripts. HR is automating workflows.

Sounds like the AI revolution is working, right?

The Gap Nobody Measured Until Now

Here's where it gets uncomfortable.

Among monthly active users, defined as people who logged in at least once in the past 30 days, 19% have never tried the data analysis feature. 14% have never used reasoning capabilities. 12% have never used search.

These aren't obscure features, they're core functionality that OpenAI highlights as transformative. And nearly one in five "active" users has never touched them.

The result is a measurable canyon inside organizations:

Frontier workers send 6x more messages than the median. For specific tasks, the divide is even more dramatic: 17x for coding and 16x for data analysis.

OpenAI calls this the "frontier vs. laggard" divide. I'd call it something else: the emergence of two entirely different workforces operating in the same office.

The Shadow AI Paradox

Here's where it gets weirder.

MIT's Project NANDA found a striking disconnect: only 40% of companies have purchased official AI subscriptions. But employees in over 90% of companies regularly use personal AI tools for work.

Nearly everyone is using AI and, shockingly, most are doing it unofficially.

This "shadow AI" often delivers better ROI than formal corporate initiatives. Why? Because employees who take initiative, by signing up for personal subscriptions and experimenting on their own time, figure out integrations without waiting for IT approval. They are pulling far ahead of colleagues who wait for official guidance.

The frontier workers aren't waiting for training programs. They're building their own operating system.

Why This Is Happening

The pattern connects directly to what we covered last month: training budgets are down 8 percentage points while AI budgets explode. Companies gave up on training their existing workforce, and now we can measure the consequences.

The gap is no longer access; everyone has the same tools at their disposal. It's also not a question of capability; frontier models can do extraordinary things. The real gap is operationalization, or the ability to weave AI into how you actually work.

OpenAI's Brad Lightcap put it plainly: fully adopting AI systems requires a mindset shift and deeper integration with how work gets done. That shift is happening unevenly, and this disparity is now quantifiable.

"There are firms that still very much see these systems as a piece of software, something I can buy and give to my teams and that's kind of the end of it."

The firms treating AI as software they "deploy" are creating laggards. The firms treating it as an operating system to learn are creating frontier workers.

The Compounding Problem

The gap isn't static. It's accelerating.

Frontier firms send 2x more messages per employee than median firms. But they send 7x more through structured workflows like Custom GPTs and Projects: repeatable, scaled processes that compound over time.

BCG data cited in the report shows organizations mastering AI deployment are pulling ahead with 1.7x revenue growth and 3.6x greater shareholder returns. The productivity advantages of frontier workers translate into organizational advantages, which attract more investment and widen the gap further.

Meanwhile, the laggards are stuck in pilot purgatory. MIT found that 95% of AI projects don't scale past the pilot stage after six months. But the failure isn't the technology, it's the organizational design around it.

What This Means If You're Building AI Products

Your users are bifurcating. Design for both.

The frontier users (top 5-10%) want power, customization, and depth. They're building Custom GPTs, chaining workflows, and pushing boundaries. Give them the tools to go further.

The median users (the other 90%) need different things entirely: lower friction, clearer defaults, and guidance built into the product. If 19% of active users have never tried your core features, that's a UX problem, not a training problem.

Onboarding is the new battleground. The gap between frontier and laggard forms in the first few weeks. Products that accelerate users through the learning curve will capture disproportionate value.

Shadow AI is signal, not noise. Employees using personal AI tools are showing you what they actually want (and what corporate deployments are failing to provide). Build for their workflows, not IT's procurement checklist.

Workflow redesign > feature adoption. The winners aren't the companies using the most AI features. They're the ones redesigning how work happens with AI as a core assumption. If you're selling "add AI to your existing process," you're selling to laggards.

The Uncomfortable Question

Here's what the report doesn't say but the data implies:

If the productivity gap between frontier and median workers is 6x (and growing), what happens to the median workers?

The optimistic view: training and organizational design can close the gap.

The pessimistic view: we're watching a skills-based sorting in real-time, and some people are on the wrong side of a widening divide.

The realistic view: both are probably true, and the outcome depends entirely on whether organizations choose to invest in their people or simply replace them with the ones who figured it out on their own.

The Bottom Line

The AI divide is no longer theoretical. It's measurable, widening, and happening inside your organization right now.

The companies winning aren't the ones with the biggest AI budgets. They're the ones creating frontier workers through workflow redesign, cultural permission to experiment, and investment in human capital.

The companies losing are easy to spot: they deployed the tools, skipped the training, and are now watching a small minority of employees capture all the value while the majority barely touches the features they're paying for.

The 6x gap isn't a problem for next year. It's already here. The only question is which side of it your organization, and your products, are designed for.

In motion,
Justin Wright

If frontier workers are 6-17x more productive than their median peers using the exact same tools, are we underestimating how much of AI's value is locked behind human factors we haven't figured out how to scale?

Food for Thought

I am excited to officially announce the launch of my podcast Mostly Humans: An AI and business podcast for everyone!

Episodes can be found below - please like, subscribe, and comment!