Beyond ChatGPT: What’s Next for Generative AI?

Generative AI is changing the role machines play in our work, completing previously time-intensive tasks like drafting reports, summarizing research, and generating code in seconds. This shift is already visible in how professionals approach everyday responsibilities. What began as an advance in text generation is beginning to reset expectations for how quickly and effectively knowledge can be produced. Yet the significance of this moment lies not only in what these systems can do, but in what they are beginning to enable.

From Generation to Agency

New workflows and systems are being built on the foundation of generative AI. Its ability to synthesize information and extend ideas at scale is transforming platforms like Claude, Gemini, and ChatGPT from simple writing assistants into powerful collaborators in ideation and critique. However, its limitations are equally clear. Because generative AI relies on historical data and produces hallucinated outputs, human judgment is vital in any meaningful application. More fundamentally, these systems stop at generation. They primarily produce insights, but they do not carry them forward independently.

Agentic AI is emerging as a response to this gap by pursuing discrete goals. Agentic models still rely on generative AI to reason and communicate, but they extend those capabilities into action, marking a shift from tools that assist with thinking to systems that participate in execution.

By introducing the capacity to plan, execute tasks, and iterate toward defined objectives, agentic AI does much more than respond to prompts. These systems can begin to navigate workflows, interact with software environments, and adjust their approach based on intermediate results.

This distinction is changing the nature of human involvement, shifting the role of the user from requester to supervisor, laying the groundwork for more integrated systems that can operate across tasks, tools, and forms of data. 

Multimodal Systems and Specialized Intelligence

Alongside the rise of agentic behavior is the expansion of multimodal capabilities that integrate text, images, audio, and video into unified systems. This advance reflects the structure of real-world problems, which rarely exist in isolation. For example, a medical diagnosis may involve imaging and narrative history, while engineering design often combines visual models with technical documentation.

At the same time, a divergence is emerging between general-purpose AI and domain-specific systems for fields like healthcare, law, and finance. Computer science practitioners are no longer tasked solely with building systems from first principles. Instead, they are increasingly responsible for orchestrating AI models, evaluating their outputs, and integrating them into larger workflows. The skill set is increasingly shifting toward system-level thinking, where understanding how components interact becomes as important as developing the components themselves.

This has direct implications for the workforce. Computer scientists should start considering how agentic and multimodal systems behave in dynamic environments, where outputs are not static but evolve through interaction. Issues of reliability and alignment are becoming more important as systems are given more autonomy. Professionals should be able to supervise intelligent systems, diagnose failure points, and ensure that automated processes remain consistent with human intent.

At UoPeople, we are shifting how computer science students are trained, teaching them to engage with AI not simply as users, but as designers and evaluators of intelligent systems. Coursework increasingly emphasizes applied experience with AI tools, ethical considerations, and the integration of AI into software development workflows. Students learn how to assess model outputs, understand system limitations, and design solutions that incorporate both human and machine capabilities.

The result is a graduate who is better prepared not only to work with existing technologies, but to adapt as those technologies evolve. As generative, agentic, and multimodal systems continue to converge, the ability to think critically about their use will distinguish professionals who can build responsibly from those who rely on them uncritically.

The Enduring Role of Generative AI

Despite the recent attention being given to agentic systems, generative AI is not receding. As agentic AI becomes more capable, it will increasingly depend on high-quality generation to function effectively. Planning, reasoning, and communication all rely on the same underlying mechanisms that power today’s generative tools.

This reflects a broader shift in how expertise is defined. When AI systems can draft reports, generate code, and summarize complex material, human contribution moves toward interpretation and decision-making. AI will continue to enhance productivity and expand creative capacity, but it will still require human oversight. The ability to ask the right questions, evaluate outputs critically, and guide systems toward meaningful outcomes is becoming more valuable than the ability to produce content alone.

For institutions and organizations, the implication is clear. Preparing individuals to work effectively with AI requires not only technical familiarity but also judgment, adaptability, and a clear understanding of where human insight must remain central. Generative AI will continue to evolve, but its long-term impact will be shaped by how thoughtfully it is integrated into human decision-making.