Analyzing how ChatGPT is being adopted across different industries, shows uneven adoption; sectors such as Information Technology, Professional, scientific & technical services, and Manufacturing leading the ranking. A distance are sectors such as healthcare, retail, construction, and transportation. This suggests that AI progresses faster in sectors with many technical profiles and fewer regulatory barriers compared to those where operational tasks or regulatory constraints prevail, as happens for example in healthcare, which, although data-intensive, is adopting ChatGPT more slowly, also due to data-privacy issues.
As for how different departments use ChatGPT in the first 90 days, it is clear that four activities dominate across the board: writing, research (factual & how-to), programming, and data analysis. But different usage mixes can be distinguished within the various departments. Technical teams (engineers, analytics, IT) strongly focus on coding and analysis; go-to-market and support roles (marketing, communications, HR) rely more on writing and content generation; designers make surprising use of both programming (for prototypes or snippets) and media generation and project managers, acting as a hub between departments, use ChatGPT in various ways, generating texts, workflows, and media. This means that, although the core functions of ChatGPT remain consistent across departments, each department adapts them to its own workflows depending on its needs.
In addition, the OpenAI report highlights that the more advanced functionalities (complex reasoning, autonomous agents, multi-document synthesis) are still little used except by technical teams.
By combining the data from the chart and the table, a coherent and layered picture emerges. Sectors with a high density of knowledge workers (e.g., engineering, analytics, IT) are more likely to experiment with advanced functionalities, transforming isolated use cases into integrated workflows, compared to sectors where the adoption of ChatGPT is slower. In the latter cases, the use of ChatGPT in complex applications is hindered by not only regulatory and governance barriers, but also organisational culture, user acceptance, and the fact that the greatest advantages still mainly come from basic uses (e.g. writing and research). This suggests that adoption depends not only on personal and cultural readiness, but also on formal rules. Without acceptance, even permissive regulatory environments may not lead to advanced usage.
Moreover, we can see that the use of ChatGPT is breaking down the barriers to skills traditionally confined to specialists. For instance, programming is becoming accessible to designers and other non-technical roles, and effective writing is now within reach of a wider range of employees beyond dedicated content teams.
However, broader challenges remain: a growing polarization between basic and advanced users, and the risk that valuable applications in sensitive areas (such as HR, finance, legal, and healthcare) may be slowed down by regulatory uncertainties and questions of trust. These are not sector-specific hurdles but cross-border issues that reflect the need for shared standards and safeguards that can support responsible adoption.
In summary, the numbers show that AI is already pervasive and empowering, but the real leap is not automatic: it requires upskilling in technical skills, regulatory policies that enable controlled experimentation in sensitive sectors, and mechanisms of assurance and transparency for the use of AI.
Source: OpenAI