Event Recap
AI tools are evolving rapidly. How can we find ways to use them that suit our own working contexts amid this constantly changing technological tide? In educational settings burdened by administrative tasks and where systems are still being established, how can AI help uncover opportunities to improve efficiency?
On 17 June, the second Cross-Domain Salon, themed “AI Collaborative Applications,” invited Song Hao-lian (A-Tai), Senior Project Manager at the School Bachelor's Office of National Chengchi University, and Lin Ching-yao, a faculty member at the College of Innovation and Design at National Taiwan University, to respond to these questions from the perspectives of administrative process automation and curriculum design. The former used practical scenarios from the School Bachelor's Office to demonstrate how Custom GPT, Vibe Coding, and Claude report organisation can systematise cumbersome administrative processes. The latter drew on an entire semester of an AI-Empowered PBL course to share how design thinking can be deeply integrated with AI tools, guiding students from passively receiving knowledge toward actively defining problems and iteratively validating solutions.
“People determine the standards and quality; AI handles the rest.”
The first speaker, A-Tai, began by candidly acknowledging that the School Bachelor's Office operates in a rather “chaotic” environment. The two cohorts together comprise nearly fifty students, there are at least two cross-unit meetings each week, and responsibilities range from application review and changes to learning plans to event organisation. Nearly all procedures are still being designed and implemented simultaneously, with almost no existing examples to follow. Under these high-pressure conditions, the team developed a “10-80-10” logic for AI collaboration: people invest 10% of their effort at the outset to establish quality standards, AI handles the middle 80% of data organisation and argument generation, and people then invest another 10% in reviewing the results.
In practice, the office primarily uses a combination of three AI tools. The first is a customised consultation chatbot built on Chat GPT (Custom GPT). Through detailed prompt design, the chatbot allows students in the School Bachelor's programme to filter out many repetitive questions through AI before submitting applications or writing learning plans, while built-in evaluation criteria guide them in improving the quality of their first drafts. With AI serving as the “first-line consultation window,” many repetitive questions can be addressed on the student side, substantially reducing the communication burden on administrative staff. The second combines Google Apps Script with Google Docs and Google Sheets to automate the sending of large numbers of personalised emails—from transcript organisation and document tab management to batch delivery of notification emails based on recipient lists. All of this is done through “Vibe Coding,” asking AI to generate the code directly and lowering the technical barrier. The third uses Claude to process lengthy reports, combining Typeless voice input and Memo AI audio transcription to create a complete workflow of “voice → transcript → AI organisation → final report.” He emphasised that the most important shift in mindset when introducing AI is recognising that “how to effectively assign AI a task” is itself a skill requiring deliberate practice. The office staff also gradually found their own rhythms through exploration. “People determine the standards and quality; AI handles the rest”—this sentence is both their work philosophy and the prerequisite for the entire process to function.
Integrating AI into a Design Thinking Course: The Double Diamond Plus “AI”
The second speaker, Lin Ching-yao, shared the “AI-Empowered PBL” course he designed at the College of Innovation and Design, embedding AI tools throughout an entire semester into the Double Diamond design thinking process (Discover, Define, Develop, Deliver, Test, Reflect). In his presentation, he used two real student projects to concretely illustrate AI’s different roles at each stage.
In the Discover stage, students used Gemini to process interview transcripts and quickly extract key insights. The instructor particularly emphasised the value of AI as an “objective third party,” arguing that, compared with researchers who can be influenced by preconceived biases, AI can identify overlooked content in interviews more neutrally. In the Define stage, AI was used to create personas and support ideation within the HMW (How Might We) framework, narrowing scattered qualitative data into concrete problem statements. The Develop stage involved the greatest degree of AI intervention: students used Gemini Canvas for collaborative design and Google Apps Script to automate questionnaire workflows. More importantly, they introduced “Vibe Coding.” Students without an information-technology background only needed to describe their requirements in natural language to generate functional interactive prototypes through tools such as Claude or Antigravity, freeing cognitive resources from “debugging syntax” for “user experience design.” In the Deliver stage, AI played a role in polishing and integration: helping organise presentation logic and refine exhibition copy, transforming rough research processes into smooth presentations of results.
For example, one group focusing on craft heritage ultimately developed a game platform called “Kinmen Flower-Drumming,” using an AI motion-correction system to help players learn traditional folk performance steps. Another group of students focused on the phenomenon of “fake rest” and created a relaxation-device prototype combining AI fatigue diagnosis with an immersive VR experience, recruiting users for testing and continuously iterating on the basis of their feedback. During the final reflection, the instructor asked each student to hand-draw “their own AI diamond” to depict AI’s position and proportion throughout the project process. Everyone drew a different shape. Some saw AI as a support layer interspersed between the two diamonds; some felt that AI enveloped the entire process from beginning to end; and some observed that AI’s degree of involvement increased as the project approached the testing and validation stage. The instructor did not rush to provide a standard answer, instead suggesting: “The lack of a definitive conclusion is what makes the discussion most valuable.”
Conclusion
During the open discussion in the second half of the event, Huang Shu-wei, a teacher at NTU’s College of Innovation and Design and the event’s moderator, raised the question of payment for AI tools. A-Tai said that students often find their own solutions and that, compared with event expenses such as venue fees, subscription costs do not in fact pose a major obstacle. Instructor Lin added that course design at the university prioritises free tools as an introduction, with greater emphasis on teaching students the “philosophy of using AI”—that is, reflecting on their relationship with tools and how to ask better questions. After all, tools will keep changing, but the ability to ask questions will not become outdated. A-Tai also shared two observations from the administrative front line. First, in recent years it has become increasingly difficult to recruit staff in administrative and educational settings, and many colleagues take on multiple roles. AI collaboration is no longer an optional enhancement; the reality is a configuration of “one person plus one AI,” and managers must recognise this shift early. Second, when colleagues familiar with the work leave, organisational knowledge is often severed along with them. His view is that comprehensive documentation is itself a prerequisite for AI automation: only with well-named and well-structured files can AI truly do its job. In other words, recording things clearly is not merely an administrative habit, but a basic condition for enabling AI to take over work.
Note: The Cross-Domain Salon is jointly organised by the Taiwan Association for Innovative Interdisciplinary Education (AITE) and the Office of the Ministry of Education’s “Interdisciplinary Flexible Study Pilot Programme.” It is expected to be held once a month, continuing to invite practitioners, researchers, and education professionals to exchange ideas and engage in dialogue.