Enterprise AI in 2025: How Artificial Intelligence Is Changing Business (OpenAI Report)
TL;DR: According to OpenAI’s 2025 report, enterprise AI is moving from the experimentation stage into production mode. ChatGPT Enterprise has grown 9x year over year. Leading companies are showing 1.7x revenue growth and 3.6x shareholder ROI compared with laggards. The gap between “frontier” and average users continues to widen.
Contents
What the OpenAI Enterprise AI 2025 report is
In 2025, OpenAI published its first large-scale report “The State of Enterprise AI”, based on two key data sources: anonymized usage statistics from more than 1 million enterprise customers and a survey of 9,000 employees from nearly 100 enterprises.
This is not a marketing document — it is an attempt to capture real patterns of AI adoption in organizations where requirements for accuracy are high, workflows are complex, and productivity improvements directly affect financial results.
The main takeaway can be stated as follows: enterprise AI is no longer an experiment. It is becoming foundational infrastructure — on par with CRM, ERP, and cloud services. The history of technological revolutions — from steam engines to semiconductors — shows that significant economic value is created precisely when companies move from learning the technology to scaling specific use cases. Enterprise AI is now entering that phase.
Key figures: scale and growth speed
The report’s data gives a clear picture of the pace of change:
ChatGPT Enterprise platform growth
| Metric | Value |
|---|---|
| ChatGPT enterprise seats | more than 7 million |
| Enterprise seat growth year over year | ~9x |
| Growth in weekly messages since November 2024 | ~8x |
| Growth in average number of messages per employee | +30% |
| Growth in weekly users of Custom GPTs and Projects | ~19x |
API and developer tools
- More than 9,000 organizations processed over 10 billion tokens
- Nearly 200 companies exceeded 1 trillion tokens
- Average consumption of reasoning tokens per organization increased by 320x over 12 months
- The number of weekly active Codex users doubled in 6 weeks
- The volume of Codex messages grew by approximately 50% over the same period
The 320x figure for reasoning tokens is especially telling: it means companies are not just using AI more often — they are using more powerful models for more complex tasks.
What employees themselves are saying
A survey of 9,000 workers from nearly 100 corporations revealed concrete measurable improvements, not just positive feelings.
Time savings
75% of respondents said their speed or quality of work improved. On average, ChatGPT Enterprise users save 40–60 minutes during an active workday. Employees in data science, engineering, and communications save even more — 60–80 minutes per day.
Expanded capabilities
75% of employees reported that they can now perform tasks that were previously out of reach for them:
- Coding support and code review
- Spreadsheet analysis and automation in Excel
- Development of technical tools and debugging
- Creating custom GPTs and agents
Growth by function
Here is how much efficiency improved in specific departments:
- 87% of IT specialists report faster resolution of technical issues
- 85% of marketers and product managers report faster campaign execution
- 75% for HR professionals — improving employee engagement
- 73% for engineers — faster code delivery
An unexpected trend: code outside IT departments
One of the most interesting findings: coding tasks are extending beyond technical roles. Over the past 6 months, the number of programming-related messages outside engineering, IT, and research has increased by an average of 36%. Marketers, finance professionals, and operations managers are increasingly writing scripts, analyzing data with code, and automating routine work.
Industry leaders and geographic reach
Growth by industry
The median sector grew by more than 6x year over year. Even the slowest-growing sector posted growth of more than 2x.
Top 3 fastest-growing industries:
- Technology — 11x growth year over year
- Healthcare — 8x growth year over year
- Manufacturing — 7x growth year over year
Financial services and professional services operate at the largest scale (by absolute message volume), while healthcare and manufacturing started from a smaller base but are now rapidly gaining momentum.
What different industries use
Technology companies focus on in-app assistants, Codex, and development automation. Financial organizations begin with customer support (with proven ROI), then move on to coding for trading, risk, and compliance systems. Professional services concentrate API spending on development and customization tools.
Global reach
International market growth is accelerating. Among the largest markets, Australia, Brazil, the Netherlands, and France are showing the fastest growth in enterprise customers — more than 143% year over year.
Specific growth figures for the number of paying business customers (November 2024 — November 2025):
| Country | Growth |
|---|---|
| Australia | +187% |
| Brazil | +161% |
| the Netherlands | +153% |
| France | +146% |
| Canada | +144% |
| Global average | +143% |
| United States | +142% |
| Germany | +138% |
| United Kingdom | +133% |
| Japan | +130% |
International growth in API customers has exceeded 70% over the past 6 months. Japan ranks first in the number of enterprise API customers outside the United States.
The gap between leaders and laggards
One of the report’s key findings is the widening inequality in AI adoption within organizations and across them.
At the employee level
“Frontier” users (95th percentile by usage intensity) generate 6x more messages than the median employee. In data analytics, the gap is even wider — frontier users apply data analysis tools 16x more often.
The gap between frontier and median users by specific task:
| Task type | Gap (frontier/median) |
|---|---|
| Coding | 17x |
| Writing and communication | 11x |
| Analysis and calculations | 10x |
| Information search | 9x |
| Practical guides | 9x |
| Creative media | 8x |
Key dependency: users working with ~7 task types report 5x greater time savings than those using only ~4 task types.
At the company level
Among active ChatGPT Enterprise users 19% have never used data analysis tools, 14% — reasoning tools, 12% — search. This indicates enormous untapped potential.
Frontier firms (95th percentile) generate 2x more messages per seat than the median enterprise, and 7x more messages through GPTs — indicating significantly deeper organizational integration.
Real-world cases: ROI of large companies
The report contains six detailed case studies. Here are the most notable results:
Intercom: Fin Voice
The company used the OpenAI Realtime API for a voice AI customer support agent:
- Reduction in response latency of 48%
- 53% of calls are resolved completely without human involvement
- Calls requiring an operator are resolved 40% faster after pre-processing by the voice agent
- Potential annual savings — hundreds of millions of dollars
Lowe's: Mylow and Mylow Companion
The home improvement retail chain deployed an AI assistant for customers and employees:
- Nearly 1 million questions per month from customers and staff
- Conversion among shoppers interacting with Mylow doubles
- Customer satisfaction index increased by 200 basis points when employees use Mylow Companion
Indeed: AI recruiting
The job platform launched GPT-powered explanations of candidate-job fit:
- Increase in started applications by 20%
- Improvement in downstream outcomes (interviews and hiring) by 13%
- Career Scout users find and apply to jobs 7x faster
- 38% higher likelihood of being hired through Career Scout
- 84% of users rate the tool as valuable
BBVA: Legal chatbot
The bank automated the verification of signatories' authority:
- More than 9,000 requests automated annually
- Freed up 3 full-time employees of lawyers
- More than 11,000 bastanteos per year
- 26% of the legal department's annual savings KPI
Moderna: Target Product Profile development
The pharmaceutical company shortened a key analytical stage:
- A process that previously took several weeks in some cases was reduced to a few hours
Financial performance of leaders (BCG, 2025)
A 2025 Boston Consulting Group study showed that over the past three years, companies leading in AI adoption achieved:
- 1.7x revenue growth compared with laggards
- 3.6x growth in total shareholder return (TSR)
- 1.6x higher EBIT margin
- Outperformance in non-financial metrics: patent activity, employee satisfaction
What leading companies do differently
The report highlights five durable patterns characteristic of organizations with the most mature AI adoption:
1. Deep system integration through context
Leaders connect connectors, giving AI secure access to corporate data within key tools. This enables context-aware responses and automated actions. Notably, roughly one in four enterprises has still not taken this step.
2. Standardization and reuse of workflows
They are actively creating, distributing, and cataloging solutions for repetitive tasks. Custom GPTs are often the foundation of this work — BBVA regularly uses more than 4,000 GPTs, which indicates a mature culture of reuse.
3. Leadership and executive sponsorship
Clear mandates, resources, team alignment, and room for experimentation — without these, scaling is impossible.
4. Data readiness and continuous evaluation
Leaders translate corporate knowledge into machine-readable procedures, build APIs for key data, and launch continuous evaluations to track real model performance.
5. Purposeful change management
They build structures that accelerate organizational learning: centralized management and training combined with decentralized dissemination through “AI champions” within business units.
Critical note from the report: OpenAI releases a new feature or capability approximately every three days. The main constraints for organizations are no longer model performance or tools — they are organizational readiness.
FAQ: Frequently Asked Questions
How many companies use OpenAI for business?
According to the OpenAI 2025 report, more than 1 million business customers and more than 7 million corporate ChatGPT Enterprise seats use the platform. Over the year, the number increased by roughly 9x.
How much time does AI actually save employees?
According to a survey of 9,000 employees, ChatGPT Enterprise users save on average 40–60 minutes on an active workday. Data science and engineering professionals report saving 60–80 minutes daily.
Which industries are adopting enterprise AI the fastest?
The fastest growth in 2025 is in technology (11x year over year), healthcare (8x), and manufacturing (7x). Financial and professional services operate at the largest scale, although they have grown more slowly from a higher base.
Is there measurable impact of AI on companies' financial results?
Yes. A 2025 Boston Consulting Group study showed that AI-leading companies achieved revenue growth of 1.7x, shareholder returns of 3.6x, and EBIT margin of 1.6x compared with laggards over the past three years.
Which tasks are best suited to enterprise AI?
According to the report, the greatest benefits come from: data analysis and financial calculations, writing and communications, IT support and debugging, coding and development, as well as marketing automation. Users working with 7 or more task types report 5x greater time savings than those who use only 4 types.
What is the difference between frontier and average AI users?
“Frontier” users (95th percentile) send 6x more messages than the median. The gap is especially large in coding (17x), writing (11x), and analysis (10x). The key difference is that they use multiple different tools and work across a broader range of tasks.
Conclusions
OpenAI’s 2025 enterprise AI report marks a turning point. This is no longer a story about whether to adopt AI or not — it is a story about how deeply and how quickly.
The data point to several major shifts:
Depth of use matters more than breadth of adoption. Companies that embed AI into specific workflows and give employees access to advanced tools are seeing incomparably greater returns.
The gap between leaders and laggards will widen. Today, frontier companies already generate 2x more messages per seat and 7x more activity through Custom GPTs. With each quarter, their advantage compounds.
Organizational readiness is the new bottleneck. Models are ready. Tools are available. The main question now is: does the company have the data, integrations, trained people, and culture to use all this?
AI expands, not merely replaces. 75% of employees now perform tasks that were previously out of reach for them. This is not only about efficiency — it is about new opportunities.
For companies that have not yet moved from pilots to production: the window is still open, but not forever. Organizations that start building systemic AI infrastructure today will look at competitors from the position of a significant structural advantage in two years.