Bottom line: In a new study, OpenAI shows that leading enterprises are increasingly using agentic AI for autonomous task execution rather than mere assistance.
OpenAI has published a study examining how enterprises are using agentic AI, ChatGPT and Codex. According to the findings, so-called “frontier” companies are pulling significantly ahead of the broader field in adoption.
The study presented by OpenAI, titled “From assistance to execution,” describes a shift in how enterprises engage with AI systems: away from pure assistance, where employees submit individual requests to a model, toward execution, where agentic AI systems autonomously handle multi-step tasks. The study is based on usage data and observations regarding the use of ChatGPT as well as Codex, OpenAI’s model geared toward software development.
For CTOs, the study’s central finding is that a group of companies OpenAI refers to as “Frontier Firms” is ahead of the curve in terms of adoption speed. These companies are not integrating agentic systems in isolated instances, but are embedding them into existing workflows and development processes. The study thus offers a clue as to which organizational patterns make the difference between experimental use and productive deployment.
For technical practice, it is relevant that Codex is cited as an example of the transition from assistance to execution in a development context: instead of individual code suggestions, agents there are increasingly taking on complete sub-tasks. Anyone looking to assess their own organization’s maturity level in AI integration will find guidance in the study on where in the development and operations process the shift from supporting to executing AI functions is already taking place.