OpenAI's Enterprise Report: 1 Million Business Customers, 320x API Growth

More than 1 million business customers now use OpenAI's tools, and API reasoning token consumption per organization has increased roughly 320x year-over-year, according to OpenAI's 2025 "State of Enterprise AI" report published this week. The figures mark the clearest signal yet that enterprise AI has moved from pilot projects into production infrastructure.

The stakes are economic, not technical. OpenAI's chief economist, Ronnie Chatterji, frames the report as evidence that "enterprise AI now appears to be entering this phase" — the phase in which, as with steam engines and semiconductors, general purpose technologies create significant value only after firms translate capabilities into scaled use cases. OpenAI also states plainly that enterprise revenue "can help fund broad, free access to powerful AI for hundreds of millions of people worldwide," tying the business segment directly to its consumer mission.

The numbers behind the shift

The report draws on de-identified, aggregated usage data from OpenAI's enterprise base plus a survey of 9,000 workers across nearly 100 enterprises. No OpenAI employee reviewed individual customer data, the company says.

The headline growth figures:

• ChatGPT Enterprise seats increased approximately 9x year-over-year; OpenAI now serves more than 7 million ChatGPT workplace seats.
• Weekly Enterprise messages grew approximately 8x since November 2024, with the average worker sending 30% more messages.
• Weekly users of Custom GPTs and Projects increased roughly 19x year-to-date, and about 20% of all Enterprise messages now run through a Custom GPT or Project.
• More than 9,000 organizations have processed over 10 billion tokens through the API; nearly 200 have exceeded 1 trillion.

Spanish bank BBVA illustrates the GPT trend: it regularly uses more than 4,000 Custom GPTs, treating AI workflows as persistent operational tools rather than one-off experiments.

Workers report real time savings

Seventy-five percent of surveyed workers say AI has improved the speed or quality of their output. On average, ChatGPT Enterprise users attribute 40–60 minutes of saved time per active day, with data science, engineering, and communications workers saving 60–80 minutes. Function-level results include 87% of IT workers reporting faster issue resolution, 85% of marketing and product users reporting faster campaign execution, and 73% of engineers reporting faster code delivery.

The more striking finding concerns task expansion. Seventy-five percent of workers report completing tasks they previously could not perform — programming support, code review, spreadsheet automation, custom GPT design. Coding-related messages have grown across all functions, and outside engineering, IT, and research, they have increased an average of 36% over the past six months. Non-technical teams are increasingly doing work that was confined to specialists.

Growth is global and broad-based

The median sector grew more than 6x year-over-year in customers; even the slowest-growing sector exceeded 2x. Technology leads at 11x, followed by healthcare at 8x and manufacturing at 7x. Professional services, finance, and technology remain the largest markets in absolute terms.

International growth is accelerating: Australia, Brazil, the Netherlands, and France grew more than 143% year-over-year in business customers, the UK and Germany rank among the largest ChatGPT Enterprise markets outside the US, and Japan has the largest number of corporate API customers outside the US. International API customer growth exceeded 70% in the last six months.

The widening divide

The report's most analytically interesting finding is the gap between leaders and laggards. Frontier workers — those in the 95th percentile of adoption intensity — send 6x more messages than the median worker, and 17x more coding-related messages. Even within data analytics roles, frontier workers use the data-analysis tool 16x more than the median. At the firm level, frontier companies generate roughly 2x more messages per seat and 7x more messages to GPTs than the median enterprise.

Depth of use correlates directly with benefit. Users engaging across roughly seven task types report five times more time saved than those using about four. Yet even among monthly active users, 19% have never used data analysis, 14% have never used reasoning, and 12% have never used search.

OpenAI identifies the constraint: the company now releases a new feature roughly every three days, and "the primary constraints for organizations are no longer model performance or tooling, but rather organizational readiness." Leading firms, per the report, share five practices — deep system integration through connectors (roughly one in four enterprises still has not enabled them), workflow standardization, executive sponsorship, data readiness and evaluations, and deliberate change management.

Case studies with hard numbers

Six case studies anchor the business-impact claims:

Intercom built Fin Voice on OpenAI's Realtime API for phone support. Latency has decreased 48% since March, Fin Voice resolves 53% of calls end-to-end on average, and calls requiring human agents are resolved 40% faster. OpenAI says Fin is "already saving customers hundreds of millions of dollars annually," given that human-handled support conversations typically cost $5–$20.

Lowe's deployed Mylow on Lowes.com and Mylow Companion to associates across more than 1,700 stores. The tools answer nearly 1 million questions per month; conversion rates more than double when customers engage with Mylow online, and customer satisfaction scores increase 200 basis points when associates use Mylow Companion.

Indeed reports that Invite to Apply with LLM-generated explanations increased started applications by 20% and improved downstream hiring outcomes by 13%. Job seekers using Career Scout find and apply to relevant jobs 7x faster and are 38% more likely to be hired.

BBVA in Mexico built a legal chatbot on ChatGPT Enterprise that automates more than 9,000 signatory-authority queries annually, redeployed the equivalent of 3 FTEs, and delivered 26% of the Legal Services division's annual savings KPI.

Oscar Health deployed member-facing chatbots integrated with claims, medical records, and service data. The platform answers 58% of benefits questions instantly and handles 39% of benefits messages with no human escalation.

Moderna used ChatGPT Enterprise to compress Target Product Profile development, reducing a core analytical step from weeks to hours in some cases. TPPs can require processing evidence packs of up to 300 pages.

OpenAI also cites a 2025 Boston Consulting Group study: over the past three years, AI leaders achieved 1.7x revenue growth, 3.6x greater total shareholder return, and 1.6x EBIT margin relative to peers — correlation, not proven causation, as the report acknowledges.

What comes next

Chatterji's forward framing is explicit: "the next phase of enterprise AI will be shaped by stronger performance on economically valuable tasks, better understanding of organizational context, and a shift from asking models for outputs to delegating complex, multi-step workflows." For competitors and buyers alike, the report's subtext is that the differentiator is no longer access to capable models — it is organizational capacity to embed them, and the divide between firms that do and firms that don't is compounding.