The era of artificial intelligence (AI) experimentation in business is officially over.
new report The State of Enterprise AI 2025, released by OpenAI on Monday (December 8), does more than simply document growth. This marks a tectonic shift in the global financial and digital economy, where generative AI is moving from an auxiliary productivity tool to an embedded, mission-critical infrastructure.
Organizations are moving beyond pilot programs to large-scale production deployments, effectively codifying their internal knowledge into intelligent, persistent assistants.
The average enterprise employee now sends 30% more ChatGPT messages per week compared to a year ago, while consumption of API inference tokens per organization has increased 320x, demonstrating both widespread adoption and deeper integration across business functions. More than 7 million workplaces now use ChatGPT Enterprise, a 9x increase year-over-year.
Implementation with ROI in mind
Adoption curves vary widely depending on readiness and resources. Large companies employing 1,000 or more people are twice as likely to build custom GPTs than smaller organizations, reflecting their greater technical ability to standardize AI-driven tasks, according to the report. Usage of custom GPT and projects has increased 19x this year, and these tools now process approximately 20% of all enterprise messages as companies codify organizational knowledge into persistent assistants.
Return on investment metrics suggest that the technology is delivering measurable value. In one study, the percentage of CFOs reporting a highly positive ROI from generative AI jumped from 27% to 85%. PYMNTS Intelligence Survey. Three-quarters of those executives are now implementing AI in their cybersecurity management.
Advertisement: SCROLL TO CONTINUE
According to Open AI reportcorporate employees believe that using AI saves 40-60 minutes per day, with data science, engineering, and communications roles reporting the highest gains at 60-80 minutes per day of activity. According to survey data from nearly 100 companies, 87% of IT professionals say faster issue resolution, 85% of marketing teams say faster campaign execution, and 73% of engineers say faster code delivery cycles.
Enterprise AI adoption trends
According to OpenAI's report, 70% of enterprise AI activities now take place within projects, which are configurable workspaces that support multi-step tasks with custom instructions and an internal knowledge base. This change reflects deeper operational use as some organizations perform AI development at scale. For example, BBVA maintains over 4,000 active GPTs. This shows that AI-driven workflows are becoming embedded infrastructure rather than supplemental productivity tools.
PYMNTS data The demand side shows the same pattern. Workflow optimization is the area with the strongest consensus among product leaders, with 98% expecting generative AI to improve internal processes, up from 70% a year ago.
Technology companies are leading the way in the use of APIs, with usage of the interfaces increasing five times year-over-year as they extend customer-facing applications such as in-product assistants and search capabilities. Non-tech companies have seen a 5x increase in API usage, suggesting adoption has expanded beyond product integration to operational deployments in customer service and content generation, which now account for approximately 20% of API activity.
Our international expansion has accelerated over the past six months, with customer growth of over 143% year over year in Australia, Brazil, the Netherlands and France. The UK and Germany rank as the largest ChatGPT Enterprise markets outside the US by number of customers, and Japan is the top international market for enterprise API customers.
This global expansion comes as the foundation model market is experiencing its most rapid changes in years. According to Menlo VenturesAnthropic now captures 40% of enterprise large-scale language model (LLM) spending, up 24% last year and 12% in 2023, overtaking OpenAI to become the enterprise leader. OpenAI's share declined from 50% in 2023 to 27%, while Google increased its enterprise share from 7% in 2023 to 21% in 2025. Together, these three providers currently account for 88% of enterprise LLM API usage, with the remaining 12% spread across Meta's Llama, Cohere, Mistral, and the long tail of smaller models.
The change in model share is consistent with how companies are deploying AI more broadly today. According to , 76% of AI use cases are purchased rather than built in-house. Data from Menlo. Despite continued strong investment in in-house builds, off-the-shelf AI solutions are reaching production faster and demonstrating immediate value while enterprise technology stacks continue to mature.
For all of our coverage of PYMNTS AI, subscribe to our daily subscription AI Newsletter.
