OpenAI · December 2025

The State of Enterprise AI: 2025 Report

OpenAI's The State of Enterprise AI: 2025 Report (December 2025; n=9,000 workers (survey) across ~100 enterprises; plus usage data from enterprise customers, fielded 2025) found that 40–60 min of time saved per active day, as reported by enterprise ChatGPT users.

Read the original report ↗Cite2 min read · Summary updated
Frontier workers' usage gap with the median worker is widest for codingMessages per user, frontier (95th percentile) worker as a multiple of the median worker, by task type, 2025
Coding
17x
Writing and communication
11x
Analysis and calculations
10x
How-to and procedural guidance
9x
Information gathering
9x
Creative media
8x
Source: OpenAI, The State of Enterprise AI: 2025 Report, 2025, p. 14.

Key findings

  1. 01
    ChatGPT Enterprise message volume grew 8x and API reasoning token use per organization rose about 320x year over year.
  2. 02
    75% of surveyed workers say AI improved the speed or quality of their output.
  3. 03
    Frontier workers send 6x more messages than the median; frontier firms send 2x as many messages per seat.

By the numbers

40–60 minof time saved per active day, as reported by enterprise ChatGPT users
75%of surveyed workers say AI improved the speed or quality of their output
36%average growth in coding-related messages outside engineering, IT and research over six months
Many monthly users have never tried advanced ChatGPT Enterprise toolsPercent of monthly active ChatGPT Enterprise users who have never used each tool, 2025
Data analysis
19%
Reasoning
14%
Search
12%
Source: OpenAI, The State of Enterprise AI: 2025 Report, 2025, p. 15.

What it means for you Draft

For executives at $10–100M companies

These are self-reports from paying customers of the vendor, so the time-savings figure is an upper-bound signal, not an expectation. The gap between heavy and light users inside the same company is the part worth noting: rolling out licenses is not the same as getting use out of them.

Limitations

Low trust.Vendor report on its own customers; survey method thinly described; publisher sells AI. Usage telemetry is real but not representative.

About the publisher: Makes ChatGPT; studies its own customers.

Cite the original

OpenAI. "The State of Enterprise AI: 2025 Report." December 8, 2025. https://openai.com/index/the-state-of-enterprise-ai-2025-report/