OpenAI has been describing, in different publications and product notes, an evolution path that is usually summarized in five steps: models → chat → agent → creativity → company. First, the nonprofit created the foundational AI models so the scientific community could experiment with them. These models were called “GPT”, in their different versions. Shortly after, it integrated them into a chat-like interface, which it called “ChatGPT”, and recently – last fall – its new versions have begun to work in “agent” modes capable of executing workflows.
The leap towards multimodal creativity
Next, the company, which has already been converted into a for-profit entity – quite a lot of profit – advocates a leap towards creative abilities multimodal that will culminate, in a few years, in an offer with governance and controls to create organizations 100% implemented with AI. Thus, throughout 2026 we will see the capabilities of ChatGPT to “think outside the box” and investigate on different issues.
Therefore, the “AI creativity” stage we are entering is not limited to producing more polished texts, or images or videos from ideas, but aims to turn AI tools into an environment of iterative creation: ideate, generate, explore, edit, compare variants and maintain consistency in the process. Based on what OpenAI has called “multi-turn generation“, the capacity for conversational iteration on the generated content itself will be complemented with “crazy ideas”, “innovative suggestions”, “brainstorming” and a host of techniques that people use to discover new methods and knowledge.
Applications in biomedicine
The applications often cited to illustrate this phase of AI range from the spectacular to the utilitarian. In biomedicine, for example, public discourse often focuses on the “cancer cure.” The prudent formulation, however, is different: the creativity of these systems can help accelerate segments of scientific work that are intensive in reading, synthesis and planning, such as reviewing technical literature, proposing hypotheses, structuring experimental protocols, debugging analysis code or preparing reports. OpenAI has presented the approach of AI as a scientific collaborator and has described the use of these tools in research contexts and in support of clinical and administrative health teams. The practical promise is not so much to make a sudden discovery as to reduce the current long cycles of discovery.
Something similar will happen in engineering. Creativity is not the same as “imagining” bridges to make yet another video with AI that saturates YouTube, but rather accelerating the iteration between requirements, design alternatives and documentation: generation of technical reports, writing responses to tenders, preparation of schedules, preparation of checklists and production of prototypes to automate calculations or prepare reports. When this layer is integrated with agent functions and enterprise environments, the value shifts toward controlled execution: teams that produce more useful drafts, faster, with traceability and access policies. OpenAI, in fact, positions enterprise adoption around security and privacy, in addition to offering integrated tools for the job.
From assistant to artifact factory
In short, the “creativity” phase can be read as the bridge between a conversational assistant and an artifact factory: a system that not only answers, but also helps to conceive, materialize and refine results, and that prepares the ground for its disciplined integration into the company. And of course, OpenAI is illustrated here just as an example of what all the big AI vendors are doing. Definitely, we are increasingly moving towards a world in which a great central intelligence will progressively replace that of our minds. To better understand the future that AI holds for us, I invite you to read my book “An infinite mind” (Tusquets editors).
