Using AI for Analysis and Research
I’m happy to talk about the use of AI in the analysis that I submitted to the CSCU Board of Regents at the end of June. Local media coverage focused on my use of AI, not the findings and recommendations themselves. I want to be clear that the narrative and organizational shape of the analysis, and the final recommendations are mine.
This analysis is not an academic paper meant for publication. It is meant to provide information, create context, and detail national best practices to help the CSCU system, and the state, make decisions. You can download it here.
From August 2025 to April 2026, I met every few weeks with then-CSCU BOR Chair Marty Guay, Vice Chair Juanita James, and Regent Ira Bloom. I was clear with them from the beginning that I would utilize multiple AI tools to aggregate and synthesize great volumes of data and resources.
My role was to direct scope, review outputs, and curate what was relevant to Connecticut. I was clear with them: it would be incredibly hard for one person to research all 50 states across these four areas in one year's time without using AI. So I would use my decades of experience, my knowledge of CT, and my understanding of the “why” behind these focus areas to soak up and carve through all of the information, and create an analysis that would “inform Board deliberations and long-term planning.”
The board members and I reviewed nearly 2,000 pages of reports and documentation, asked questions and battle-tested lenses and approaches to understanding this work; and fine-tuned the analysis and recommendations together.
In each area of workforce, funding, and early college, I generated individual state reports, then rolled them into separate regional subject area reports. I studied each report, noting significant details, histories, programs, frameworks, and factors that were salient to Connecticut. For the system office structure analysis, I determined a set of peer systems, and then reviewed appropriate metrics across the systems in comparison to CSCU.
I used AI to provide additional analysis if there were questions or gaps, particularly in areas in which I do not have first-hand experience or technical expertise. Throughout the process, AI was used through multiple rounds of editing and revision to copyedit and scrutinize the content. As a final product began to take shape, additional checks were run to specifically detect inconsistencies, fallacies, and hallucinations.
I also had nearly two dozen phone and in-person conversations with higher ed leaders from across the country. The purpose of these candid conversations was to discuss particular programs, structures, and processes. Their personal insights tested the veracity of particular findings against the experiences of leaders and practitioners on the ground.
In coming posts, I will detail findings in each area, and how I believe AI can help those who do this kind of work produce deliverables of value.
Comments