LearnAI: Just-in-Time AI Co-Creation Across Disciplines at a University

In the modern world, artificial intelligence has moved from science fiction to a daily utility, reshaping how people solve problems in offices, classrooms, and homes. Yet, a significant gap remains in how society learns to use these tools. Most educational efforts fall into one of two camps: broad, theoretical workshops that explain what AI is without teaching how to use it, or advanced technical courses designed for computer science experts that leave everyone else behind. This divide leaves a vast group of learners—those who need to build a website, automate a spreadsheet, or organize a workflow but lack coding skills—without a path forward. They face a choice between doing everything manually or handing over their work entirely to a machine, often without understanding the process. The question facing educators today is not just how to teach people to code, but how to help diverse groups of people collaborate with AI to build real, working solutions that fit their specific needs.

At Monmouth University, a team of researchers and educators tested a new approach to bridge this gap, creating a program called LearnAI. Instead of forcing students to learn complex programming languages first, they built a service that meets people exactly where they are, offering immediate help with real-world tasks. The program operates on two levels. The first is a wide net cast across eighteen different university courses, ranging from nursing and business to computer science. In these classes, undergraduate tutors gave short, fifteen-minute talks showing how AI could solve immediate problems, such as creating a study guide or designing a portfolio website. These sessions were designed to spark interest and demystify the technology, inviting anyone curious to sign up for deeper help.

For those who took the next step, the second layer of the program began: a series of one-on-one sessions where clients worked directly with a trained student tutor. These meetings were not lectures; they were workshops where the client brought a specific goal, like building a departmental newsletter or automating a data entry task. The tutors followed a structured five-step guide to help the client turn that goal into a finished product. First, they helped the client define the problem clearly, moving past vague ideas like "I want a website" to specific requirements like "I need a site where faculty can log in to post updates." Next, they matched the right tools to the task, choosing between different AI platforms based on what the client already knew and what constraints they faced, such as limited computer access.

The core of the process involved the client and tutor working together to write instructions for the AI, a practice the researchers call "prompting as specification." Instead of asking the AI to simply "do it," the client learned to write detailed requests that acted like a blueprint. The tutor would demonstrate how to write these instructions, then gradually step back, letting the client take the lead while offering guidance. Once the AI generated a result, the pair moved to the most critical phase: verification. They tested the output to ensure it actually worked, checking for errors or features the AI might have invented that didn't exist. Finally, they discussed the ethical implications, deciding how the tool should be used, what data it could handle, and how to credit the human effort behind the creation.

Over two semesters, this model supported thirty-five clients, including students, faculty members, and administrative staff. The results were tangible and diverse. The group co-created thirty-six portfolio websites and more than twenty deployed web applications. These were not practice exercises; they were functional tools used in real life. A nursing professor built a secure system for tracking student attendance that was used by over fifty students. An education student with no coding background created a personal website and an email-refinement tool, which she later used to prepare materials for a professional conference. An administrative staff member on a restricted computer learned to automate repetitive data cleaning tasks using tools available on her locked-down desktop.

The most significant finding, however, was not just the number of websites built, but a shift in how the clients thought about the technology. Before the program, many participants viewed AI as an "oracle"—a passive machine that simply provided answers when asked. After working through the process, their perspective changed. They began to see the AI as a "process partner," a collaborative tool that required human direction, oversight, and verification. One computer science professor noted that the AI started asking her questions that made her think about angles she hadn't considered, transforming the interaction from a simple query into a genuine collaboration. Another staff member, initially overwhelmed by the technical steps, learned to break tasks down and eventually taught her colleagues how to use the same methods.

The researchers observed that this shift did not happen automatically; it required the human element of a tutor to guide the client through the frustration of learning. The tutors, who were all computer science majors, found that their most important skill was not technical expertise but the ability to communicate and adapt their teaching style to the client's comfort level. They learned to slow down for those who felt overwhelmed by the number of tools and to speed up for those who were already familiar with the technology. The program also revealed important boundaries. Some clients felt that the effort required to prompt the AI outweighed the benefits, and others chose to reject AI use entirely for certain types of work, such as scholarly writing, preferring to maintain full personal control. These reactions were not seen as failures but as valid choices that highlighted the importance of letting individuals decide when and how to use these tools.

While the program showed promising results, the researchers were careful to note the limits of their study. The data came from a small group of people who volunteered for the program, so it cannot be assumed that every student or teacher would have the same experience. The measurements of learning were preliminary and based on a small number of participants, suggesting the need for further study to confirm the long-term effects. However, the collection of real-world artifacts—the working websites and applications—provided concrete evidence that this approach works. The project demonstrated that with the right support, people from all backgrounds can move from being passive consumers of technology to active creators who use AI to solve their own problems. The LearnAI framework offers a practical model for institutions looking to integrate artificial intelligence into education not as a separate subject, but as a tool woven into the fabric of daily work and learning.