Much and more has been said about artificial intelligence (AI) in education since the release of ChatGPT 3.5 in November 2022. However, the focus is most often on the classroom with ample discussions of chatbots writing essays in English classes or multi-agent systems architecting and implementing three-tier applications in computer science classes. For community college leaders, however, AI’s impact is far less transactional and significantly more structural.
To illustrate, AI in higher education administration functions as a decision-support infrastructure that bridges institutional silos. It is the algorithm identifying a student as at risk within advising and student success systems. It is the predictive model informing scheduling, enrollment planning, and resource allocation decisions across the institution.
Community colleges often act as primary engines of social mobility in their local communities for first-generation, immigrant, and working adult students. At present, however, they find themselves at a crossroads. The introduction of AI into the community college ecosystem represents both a technological shift and a restructuring of how opportunity is distributed. To lead through this shift, community college administrators must guide the conversation away from facile conversations of efficiency metrics and toward one innately aware of the issues of ethics, power and accountability that are inherent in operating in an AI-rich world.
The Power Gap: Buying vs. Building Success
Any modern institution of higher education is enabled by an increasingly complex web of software- and infrastructure-as-a-service (SaaS/IaaS) platforms, such as Navigate360 for student success, PeopleSoft for resource planning, or Google Cloud Platform to provide IT and data services to the institution. When institutions subscribe to these platforms, they are not simply purchasing access to software or cloud infrastructure. In many cases, they are also adopting an embedded philosophy of what student success or effective resource allocation should look like. With the proliferation of AI into the enterprise space, we’re now seeing the SaaS/IaaS model expand to model- and AI-as-a-Service (MaaS/AIaaS).
In this regard, a significant structural disparity exists. While large research universities often have the resources to deploy and audit values-aligned enterprise AI models, community colleges frequently rely on commercial tools whose underlying logic may not always be fully visible to institutional stakeholders, creating an environment where software design decisions made outside the day-to-day realities of community college operations can influence which variables are prioritized in assessments of student success, employee engagement or institutional effectiveness.
This means that, if only senior IT leadership and the college’s C-suite are involved in procurement and integration discussions, we risk excluding the voices of the disparate constituencies that make up every college—advisors, faculty, and the students these systems most affect. Modern AI governance in the community college space, therefore, is not an AI issue but a mission-critical leadership responsibility.
The Mirror of History: When Data Misinterprets Resilience
A commonly overlooked aspect of AI is its retrospective nature. While users are often aware of the knowledge cutoff of modern AI systems, AI also looks to the past to identify patterns that might hold true at some point in the future. This means that, if our historical data reflects decades of system inequality in grading, financial aid, or admissions, our AI systems will not rectify these gaps on their own. Instead, they risk automating them at scale.
For community colleges, this reality is particularly dangerous. Our students often follow nontraditional paths: part-time enrollment, full-time employment, and sometimes considerable caregiving demands. Most predictive models, however, have historically been trained on data drawn disproportionately from four-year residential institutions. While retrieval-augmented generation (RAG), seeks to ameliorate this gap by grounding outputs in an institution’s own data, this method is only as good as the data being used as that ground truth. And it’s all but a truism that the data shops in many community colleges are under-resourced compared to their four-year counterparts and that their data is, therefore, of varying quality.
So, if we are not cautious, we could end up in a situation where an algorithm in one of our intelligent student success and advising systems sees a student enrolled in nine credits while working thirty hours a week, and they end up labeled as high risk, triggering a cascade of potentially unwelcome outreach. As educators, however, we may recognize that students’ resilience through more holistic engagement with that learner, which demands specialized training that makes use of intelligent systems to avoid relying on so-called algorithmic authority, which is the tendency to trust a machine’s output simply because it appears mathematically precise. Doing so can inadvertently penalize the very students we are attempting to serve and support.
A Framework for Action: The Human-in-the-Loop
Said simply, ethical accountability cannot be delegated to an API call for an intelligent decision support system. While algorithms can process data at an inhuman scale and help address some of the resource gaps in the community college space, they lack contextual nuance. To bridge this gap, I advocate for a human-in-the-loop (HITL) framework, positioning AI-based systems as collaborators rather than substitutes.
For administrators, implementing this framework means:
- Contextual auditing: ensuring all recommendations AI-based systems generate are reviewed by the appropriate community member who understands the community and its needs—faculty for curriculum and learning, administration for strategic matters, staff for operational ones.
- Vendor transparency: demanding that third-party providers explain the why behind their system’s recommendations. Institutions should be cautious about implementing recommendations that cannot be meaningfully explained or interrogated. If their AI systems aren’t transparent and explainable, we should think twice before deploying them on our data.
- Collaborative governance: including faculty and student advocates in the selection and implementation of administrative AI tools.
The Leadership Mandate
Always bear in mind that AI has the potential to reflect our existing power structures, but—if used intelligently and critically—it can also be deployed to help disrupt them. The same tools that perpetuate bias can be leveraged to reveal previously unnoticed barriers to student success or to highlight patterns of inequitable resource allocation but doing so requires human intervention and collaboration. It demands a human in the loop.
As leaders, our responsibility is not to become computer scientists but to develop AI literacy and fluency as administrators, which means asking the critical questions:
- Where did this data come from?
- Whom does this model serve, and whom does it exclude?
- Who is accountable when the model fails?
Institutions that address these questions will do more than just adapt to the AI transformation; they will lead it, ensuring that the open door remains open for everyone.