The Gap Went from 2x to 3.5x in Twelve Months. It Was Never About Using AI More.

OpenAI just released the first edition of B2B Signals, a recurring measure of how AI is actually diffusing across businesses. The headline number: frontier firms now use 3.5 times more AI intelligence per worker than typical firms. A year ago, that ratio was 2x.

That is not a steady climb. That is acceleration. The companies ahead are pulling away faster than the companies behind are catching up. And the most important finding in the entire report has nothing to do with the 3.5x number.

It’s this: message volume only explains 36% of the frontier advantage.

The other 64% comes from how these companies use AI. Not how much.

The Volume Trap

Most organizations measure AI adoption by counting usage. Seats activated. Messages sent. Queries per employee per week. The assumption is that more activity equals more value, and therefore the path to catching up is getting more people to use the tools more often.

OpenAI’s data says that assumption is wrong.

The frontier firms are not just sending more messages. They are delegating complex, multi-step tasks. They are using agent-based tools that integrate with their actual workflows. They are running AI inside code reviews, research pipelines, and operational processes where the output feeds directly into the next decision.

Codex usage tells the story most clearly. Frontier firms send 16 times more Codex messages per worker than typical firms. Not 16% more. Sixteen times. That gap is not a volume difference. It is a structural difference in how work gets done. Those teams have rewired their development process around AI as a working partner. The rest are still treating it as an occasionally useful search bar.

What the 64% Actually Looks Like

When OpenAI breaks down where the gap lives, the pattern repeats across every product category. ChatGPT Agent, Apps, Deep Research, custom GPTs. In every case, frontier firms adopt the more complex tools at dramatically higher rates.

This is the part that matters for any leader making AI decisions right now. The gap is not about access. Everyone has access to the same tools at the same price. The gap is about organizational readiness to actually use those tools for real work.

Complex tools require clear workflows to be useful. Agent-based task delegation only works when a team has defined what the agent should do, what “done” looks like, and how the output connects to the next step. Custom GPTs only deliver when someone maps them to a specific role in a specific process.

The frontier firms did that work. They built the scaffolding. Policies for what AI can decide autonomously. Feedback loops for when AI gets it wrong. Integration points where AI output enters the business process instead of sitting in a separate window.

The 80% of firms on the other side of this gap have not done that work. And the gap is compounding because every week those scaffolds exist, the teams running them get better at using them. The teams without scaffolds are just running the same standalone prompts they ran last quarter.

Why Six More Months Costs More Than Six Months

The math here is uncomfortable.

If the gap went from 2x to 3.5x in twelve months, the trajectory suggests it reaches 5x or higher in the next twelve. That is not a prediction. It is the shape of a curve where one group is compounding and the other is linear.

And the frontier advantage is self-reinforcing. Teams that delegate complex tasks to AI free up capacity to find more tasks to delegate. Teams that build feedback loops make their AI better every week, which makes the next use case easier to deploy. Teams that wire agents into their real workflows generate proprietary data on what works, which competitors cannot replicate by buying the same subscription.

Goldman Sachs, Cisco, DoorDash, State Farm. The companies OpenAI names as expanding their usage are not startups experimenting. They are large enterprises that made structural decisions about where AI lives in their operations. Those decisions were made months or years ago. The B2B Signals data is the compound interest on those decisions becoming visible.

If your current plan is to “increase AI adoption” by getting more employees to use ChatGPT more often, you are solving for the 36%. The other 64% of the gap requires changing how work gets done. That takes months. It takes governance frameworks, integration work, workflow redesign, and people who know how to make AI stick in a process. Every month you delay that work, you are not standing still. You are falling behind at an accelerating rate.

Where This Leaves Most Companies

The B2B Signals data gives a precise timestamp to something a lot of leaders feel but have not been able to name. The gap is real. It is measured. And it is not closing on its own.

The companies at 3.5x did not wake up one morning and decide to be frontier firms. They built small operational habits around AI over 12 to 24 months. They made it normal to delegate, automate, and integrate. Each habit was minor. The cumulative effect is an advantage that cannot be purchased.

The 3.5x number will be higher next quarter. The question is not whether your company uses AI. It is whether the way you use it is building compound advantage or just adding volume to a workflow that has not fundamentally changed. OpenAI just gave you the data to answer that question honestly.