What the 2025 State of Enterprise AI Report Reveals About AI’s Growing Role in Business

OpenAI’s 2025 State of Enterprise AI report, compiled from more than one million anonymous enterprise customers and a survey of nearly 9,000 employees, demonstrates that AI is moving from experimental projects to core infrastructure within businesses.

1. Rapid Scale and Deeper Workflow Integration

ChatGPT Enterprise message volume grew about 8× year‑over‑year, while API token consumption per user increased roughly 320×. The number of enterprise seats surpassed 7 million, and ChatGPT Enterprise seats rose about 9×. Custom GPTs and Projects allow AI to embed company knowledge bases and connect internal systems, enabling multi‑step task automation; weekly active users of Custom GPTs grew about 19×, handling roughly 20% of enterprise messages. For example, BBVA has deployed more than 4,000 custom GPTs, creating an internal AI‑tool ecosystem.

On the developer side, API usage exploded: over 9,000 organizations processed more than 100 billion tokens, with around 200 organizations exceeding one trillion tokens. The new Codex model showed strong performance in code generation and debugging, doubling weekly active users and increasing message volume by about 50%.

2. Tangible Productivity Gains for Employees

According to the survey, 75% of employees feel AI improves their work speed or quality, saving 40‑60 minutes per active day; data‑science, engineering, and communication roles save even more, up to 60‑80 minutes. Role‑specific impacts include 87% of IT staff reporting faster issue resolution, 85% of marketing and product staff seeing quicker campaign execution, 75% of HR noting higher employee engagement, and 73% of engineers delivering code faster.

AI is also breaking traditional role boundaries: 75% of employees can now perform tasks they previously could not, such as programming, data analysis, and automation, with non‑technical coding‑related messages increasing 36% over the past six months.

Deeper usage of advanced features yields greater time savings: users who heavily leverage capabilities like deep research or image generation save more than 10 hours per week, consuming eight times the AI output of users who see no time savings.

3. Global Growth and Industry Variations

Enterprise AI adoption grew a median of 6× across industries in the past year, with the technology sector leading at 11×, followed by healthcare (8×) and manufacturing (7×). The largest deployments remain in professional services, finance, and technology.

API usage patterns differ by sector: technology companies primarily embed AI assistants and search within products, professional services favor coding tools, and finance focuses on customer‑support applications. Non‑technology firms saw API usage increase 5×.

International growth is especially strong: API customers grew over 70% in the last six months, with paid‑enterprise customers in Australia, Brazil, the Netherlands, and France expanding more than 143%, far outpacing the global average. Japan has become the largest enterprise API market outside the United States.

4. Gap Between Leaders and Laggards Widening

Employees in the top 5% of usage send six times more messages than median users and perform data‑analysis tasks 16× more often; coding‑task gaps reach up to 17×. At the company level, leading firms generate twice the messages per seat and use Custom GPTs seven times more than median firms. Nevertheless, 19% of monthly active users have never used a data‑analysis tool, and 14% have never used inference capabilities.

Case studies illustrate concrete benefits:

Intercom leveraged real‑time API to build a voice AI客服, reducing latency by 48% and achieving a 53% fully‑automated resolution rate.

Lowe’s deployed the Mylow assistant, doubling online conversion rates and improving in‑store客服 satisfaction by 200 basis points.

Indeed used AI to match candidates with jobs, raising application start rates by 20% and increasing hiring success by 13%.

BBVA’s legal AI assistant automated over 9,000 queries annually, saving the equivalent of three full‑time employees.

Oscar Health’s medical chatbot answered 39% of benefit‑related questions without human intervention.

Moderna compressed target product profile (TPP) drafting from weeks to hours using AI.

Conclusion: Depth Determines Future Success

OpenAI’s chief economist Ronnie Chatterji notes that enterprise AI will evolve toward higher performance, better contextual understanding, and more complex workflow delegation. Leading companies are already integrating data, standardizing workflows, and strengthening change‑management practices, while many firms still have a substantial catch‑up gap. For Chinese enterprises, the report serves as both a warning and an opportunity—early action can secure a competitive advantage.