Why AI matters at this scale

The Dyson Grand Challenges program is an experiential learning initiative within Cornell University's Charles H. Dyson School of Applied Economics and Management. It engages undergraduate students in year-long, team-based projects addressing complex, real-world business and societal problems. As part of a major research university with over 10,000 employees, the program operates at a scale where manual coordination of personalized learning experiences becomes increasingly burdensome. AI presents a critical lever to maintain educational quality and personalization while managing administrative complexity and growing student cohorts. For a large, resource-rich institution like Cornell, AI adoption is not about survival but about enhancing prestige, improving learning outcomes, and optimizing the use of expert faculty time. The sector is increasingly competitive in offering distinctive, applied learning, making technological innovation a strategic differentiator.

Concrete AI Opportunities with ROI Framing

1. Intelligent Student-Project Matching Engine: Manually matching hundreds of students to suitable Grand Challenges projects is time-consuming and suboptimal. An AI matching algorithm that analyzes student transcripts, expressed interests, skills, and past project success data can create higher-performing teams and increase student satisfaction. ROI is realized through improved retention, higher project success rates (enhancing the program's reputation), and freeing 40+ hours of advisor time per matching cycle. 2. AI-Enhanced Project Design Assistant: Faculty and industry partners spend significant effort scoping new challenge projects. A fine-tuned LLM can ingest past project briefs, successful outcomes, and current event data to generate draft project charters, learning objectives, and resource lists. This can cut project setup time by 30-50%, allowing the program to scale its project portfolio without proportional increases in staff. 3. Predictive Analytics for Student Support: Attrition and disengagement in long-term projects are costly in terms of lost tuition and unmet learning goals. An AI dashboard that aggregates data from LMS interactions, peer evaluations, and milestone submissions can identify students at risk of falling behind. Early, targeted intervention preserves the investment in each student and improves overall program completion metrics, directly supporting enrollment and funding goals.

Deployment Risks Specific to Large Universities

Implementation at a university of Cornell's size faces distinct hurdles. Governance and Procurement cycles are lengthy, requiring multiple committee approvals for new software, especially involving student data. Data Silos are profound; student information, learning management, and project data often reside in separate, legacy systems (e.g., PeopleSoft, Salesforce), complicating integration for a unified AI tool. Cultural Resistance from faculty concerned about AI undermining pedagogical autonomy or from staff fearing job displacement can stall pilots. FERPA Compliance imposes strict boundaries on student data usage, limiting the datasets available for training models and requiring robust privacy-by-design architectures. Finally, Funding Models are often annual and departmentally siloed, making it difficult to secure sustained investment for an interdisciplinary program's AI initiative, which may be seen as an IT cost rather than a core educational investment.