The Pulse of Higher Ed

Will artificial intelligence close the student success gap or widen it into a permanent caste system?

That is the question higher education leaders should be asking as AI advising tools move from pilot to procurement. The temptation is to treat AI as a cost-savings lever: deploy a chatbot, deflect the tickets, claim a productivity win. But the most instructive case study of the last three years points in a different direction, one where AI is deployed precisely because equity is the goal, not the casualty.

In Chapter 8 of AI Applications in Online Higher Education Administration, I trace the evolution of AI student support technologies from rule-based chatbots to context-aware AI assistants to autonomous AI agents. The question is not what AI can do, but what we choose to let it do and for whom.

The Hybrid Advising Co-Op as Proof of Concept

Founded in 2022 by the Bill & Melinda Gates Foundation and facilitated by Shift, the Hybrid Advising Co-Op brought together six organizations that share a common student population: first-generation, low-income, and historically underserved learners. The partners like Bottom Line, Let’s Get Ready, OneGoal, College Advising Corps, KIPP Public Schools, and technical partner Mainstay, each developed a distinct hybrid model.

Bottom Line’s Blu leans bot-forward, automating the high-volume transactional layer of advising. Let’s Get Ready built a near-peer coaching system where AI supports human coaches rather than substitutes for them. Across all six models, the design principle was the same: AI absorbs the predictable, repetitive load like appointment reminders, FAQ responses, deadline nudges, registration logistics, so that human advisors can invest their finite time in the relational, high-stakes mentoring that actually moves the needle for vulnerable students.

The Two-Tier Risk Is Already Here

The ethical concern Chapter 8 names explicitly is the emergence of a two-tier advising system: privileged students continue to receive high-touch human advising, while underserved students are routed to automated agents. This is not a hypothetical. It is the path of least resistance for institutions facing high caseloads, advisor turnover, and pressure to scale.

The data should make us cautious. Khan Academy’s MAP Accelerator study, what is now called the “5 Percent Problem”, found that only the small fraction of students who engaged with the platform more than 30 minutes per week showed measurable academic gains. The other 95% were effectively excluded from the reported benefit. Automated systems tend to reward students who already possess the time, infrastructure, and self-direction to use them. The students who most need advising are precisely the ones least likely to extract value from a self-service AI experience.

Add cultural competency to the equation and the risk sharpens further. AI models trained on colorblind assumptions can perpetuate bias against the students they are nominally serving. The U.S. Department of Education’s Office for Civil Rights, acting on Executive Order 14110, has already flagged discrimination risks in AI deployments affecting minority students, multilingual learners, and students with disabilities.

Chapter 8’s Five Ethical Standards

For UPCEA members designing AI-enabled advising in professional, continuing, and online divisions, divisions that disproportionately serve adult, first-generation, and underrepresented learners, Chapter 8 offers five ethical anchors:

• Autonomy. AI suggestions should never override student agency or professional judgment.
• Fairness. Equitable treatment is not a feature; it is a precondition.
• Transparency. Students should know when and how AI is shaping their advising experience.
• Accountability. Responsibility must be assignable when AI-generated advice fails.
• Human-Centered Design. Some advising scenarios require empathy that only a human can provide. Design for that, do not engineer around it.

Three Moves for UPCEA Practitioners

If your institution is implementing AI in student support, three practical moves protect against the two-tier outcome.

First, audit the routing logic. Inequity rarely lives in the algorithm itself. It lives in the decision about which students get the chatbot and which students get the appointment. Map the distribution. Look for patterns by income band, first-generation status, and program modality.

Second, automate the transactional, protect the relational. The Co-Op’s design discipline is replicable. Identify the advising tasks that are genuinely repetitive like registration FAQs, milestone reminders, document collection and let AI carry that load. Then deliberately reinvest that recovered advisor time in deeper relationships with the students whose success is most fragile.

Third, do not let efficiency refill the caseload. Chapter 8 warns that the natural institutional response to advisor capacity gains is to expand caseloads, not deepen service. If your AI deployment results in advisors carrying 400 students instead of 300, you have not solved a problem. You have entrenched one.

The same design logic that drives the Hybrid Advising Co-Op operates at the individual faculty level. In my open-access guidebook The Learn-It-All Educator (Machajewski, 2026), I argue that AI should function as a cognitive gym, not a cognitive elevator, a tool that adds productive friction rather than removing it. The Cognitive Triage framework distinguishes FLUFF (Formatting, Layouts, Under-the-hood, Filing, Filtering: work worth delegating) from SPARK (Specific, Persuasive, Authentic, Rigorous, Keen-insight: ideas worth thinking and protecting for human thought). The Co-Op essentially applied this split at the institutional scale: AI handles the FLUFF of advising (reminders, FAQs, registration logistics) so advisors can invest in the SPARK of mentorship. The principle scales from one faculty member’s workflow to a six-organization consortium. Free OER: https://doi.org/10.5281/zenodo.19041123

 

The Real Standard

The standard for AI in higher education advising is not efficiency. It is whether the deployment expands access to human relationship for students who have historically been denied it. The Hybrid Advising Co-Op met that standard. The question for each of us is whether our institutions will.

 

Dr. Szymon Machajewski is Associate Director of Academic Technology and Learning Innovation at the University of Illinois at Chicago and Adjunct CIS Professor & Fellow for Teaching with AI. Dr. Machajewski is the author of The Learn-It-All Educator: A Guidebook for Training Brains, Not Replacing Them with AI and “Chapter 8: AI in Online Administration” in AI Applications in Online Higher Education Administration: Strategies for Maximizing Returns and Improving Outcomes.