OpenAI just released their State of Enterprise AI report — 9,000 workers surveyed across ~100 enterprises, plus usage data from 1M+ business customers.
Most of the headline metrics are “message counts” which is a vanity metric. A message could be “rewrite this paragraph” or “build me a financial model.” Same count, wildly different value.
Here’s what’s actually useful for founders and B2B execs to know:
#1. 320x Increase in Reasoning Token Consumption Per Org YoY
This is the real signal. Not messages sent — actual compute consumed on frontier reasoning models. Enterprises aren’t just chatting with AI. They’re running complex, multi-step reasoning tasks in production. 320x in 12 months means this went from experiment to infrastructure.
#2. 40-60 Minutes Saved Per Worker Per Day
Data scientists and engineers report 60-80 minutes. Self-reported, yes, but directionally significant.
At scale: A 500-person company with 50% adoption is reclaiming 1,000-1,500 hours per week. At $75/hour fully-loaded, that’s $4-6M annually in productivity gains. This is the ROI math your CFO buyers need.
#3. 75% of Workers Can Now Do Tasks They Couldn’t Do Before
Not faster. New capabilities.
The report specifically calls out non-technical workers doing: code review, spreadsheet automation, technical troubleshooting, and custom tool development.
This is TAM expansion. The addressable market for “technical work” just got dramatically bigger because the definition of who can do technical work changed.
#4. 36% Growth in Coding Activity From Non-Engineering Functions (6 Months)
Sales ops writing Python. Marketing building automations. Finance creating custom analysis tools.
In six months. This isn’t a 3-year digital transformation initiative. This is happening now, bottoms-up, whether IT sanctioned it or not.
#5. Custom GPTs: 19x Growth YTD, Now 20% of Enterprise Workflows
BBVA runs 4,000+ custom GPTs in production.
This is the “internal apps” story. Enterprises are codifying institutional knowledge — playbooks, processes, tribal knowledge — into reusable AI workflows. The companies doing this are building compounding advantages. The ones that aren’t are watching their best employees do it anyway in shadow IT.
#6. 25% of Enterprises Haven’t Connected AI to Company Data
One in four organizations paying for ChatGPT Enterprise haven’t turned on the connectors. They’re running a Ferrari on an empty tank.
Meanwhile: 19% of monthly active users have never tried data analysis. 14% have never used reasoning models.
The gap isn’t access to AI. It’s organizational capability to actually use it. This is where the real work is.
The Big Takeway: The Model Isn’t The Constraint in The Enterprise (Anymore)
The report buries the lede: “The primary constraints for organizations are no longer model performance or tooling, but rather organizational readiness.”
OpenAI ships new capabilities every ~3 days. The models are good enough. The bottleneck is companies’ ability to integrate, train, and change workflows.
For B2B founders: The product opportunity isn’t “more AI features.” It’s helping enterprises actually deploy and adopt AI effectively. Integration. Training. Change management. Workflow redesign. That’s where the value gets created — and captured.
