Enterprise use of AI is shifting from small pilot projects to deep operational integrations, according to new data released by OpenAI. The company says businesses are no longer treating generative models as simple assistants for summaries or short replies. Instead, they’re assigning multi-step tasks and wiring models directly into their products and internal systems.
OpenAI now serves more than 800 million weekly users across its platform, and more than a million businesses rely on its tools. That scale is feeding a cycle in which everyday familiarity with AI is accelerating its adoption in the workplace. The report highlights a clear trend: productivity gains are real, but the gap between “frontier” users and the average organisation continues to grow.
From basic chat to complex reasoning
While the volume of ChatGPT messages has risen eightfold over the last year, OpenAI says a more telling metric is the surge in API reasoning tokens, which reflects heavy back-end usage. Consumption has grown nearly 320 times per organisation, signaling that companies are directing models to handle logic, analysis, and chained decisions.
Custom GPTs and Projects, tools designed to embed company-specific knowledge, are seeing similar momentum. Weekly usage has climbed about 19 times in 2024, and roughly one in five enterprise messages now moves through these customised environments. Standardised workflows are quickly becoming the norm for professional use.
That’s translating to measurable time savings. Workers report gaining 40–60 minutes per active day through AI tools, with data scientists, engineers, and communication teams seeing closer to 60–80 minutes. The shift is reshaping job boundaries, especially as coding tasks spread beyond technical roles. Over the past six months, coding-related queries from non-technical teams have grown by an average of 36 percent.
Ryan Daws
Operational benefits range widely. Nearly nine in ten IT workers say AI helps them resolve issues faster, and three-quarters of HR professionals report stronger employee engagement.
A widening gap in enterprise AI maturity
One of the clearest findings is the divide in how deeply organisations integrate AI. Workers in the top 5 percent of adoption generate six times more messages than the median user. Frontier firms send twice as many messages per seat and seven times more messages to custom GPTs than the typical enterprise.
The difference comes down to breadth. Employees who use AI across seven or more task types save about five times more time than those who limit themselves to a few basic functions. Light use produces limited returns; deep use changes workflows entirely.
Industries that jumped in early—professional services, finance, and technology—still lead in adoption, but others are scaling quickly. Tech usage grew 11x year-over-year, while healthcare and manufacturing followed with 8x and 7x growth. International expansion is strong as well. Markets such as Australia, Brazil, the Netherlands, and France have grown more than 140 percent since last year, and Japan now has the most corporate API customers outside the United States.
Real-world examples show how AI is reshaping operations
Several case studies illustrate the shift from experimentation to measurable business outcomes.
Lowe’s rolled out an AI tool for store associates across more than 1,700 locations. Customer satisfaction scores rose by 200 basis points when employees used the system, and online shoppers who engaged with the retailer’s AI assistant were more than twice as likely to complete a purchase.
Moderna used AI to streamline its Target Product Profile process, which usually requires weeks of cross-team work. By automating the process of extracting facts from large evidence packs, the company cut core analytical steps from weeks to hours.
BBVA applied enterprise AI to clear a bottleneck in legal validation. Automating more than 9,000 annual queries freed up the equivalent of three full-time staff members to focus on higher-value work.
BBVA
These examples underscore a central point: the biggest barriers now are organisational, not technical. About one in four enterprises still hasn’t connected models to internal systems, leaving them limited to general knowledge rather than company-specific insight. The firms making the most progress have strong executive support and invest in documenting and structuring their institutional knowledge so it can be reused at scale.
