The challenge with using general AI for learning

When Rajen Sheth published 10 Lessons on How to Drive Learning with AI, one theme stood out: educational AI should not be evaluated like general-purpose AI.

Large language models excel at retrieving information: ask a question and they generate an answer. But learning requires something different.

Research shows that approximately 75% of students do not know enough about a subject to ask the right question in the first place. If learning depends entirely on student prompts, many learners will struggle before they even begin.

Why AI should ask the questions

Traditional AI systems wait for a user to initiate a conversation.

Learning-focused AI reverses that model.

Sheth argues that effective educational AI should identify where students are struggling, ask the right questions, and guide learners toward conceptual understanding. Rather than simply delivering information, the system should function more like an instructor helping students work through ideas step by step.

The goal is not faster answers. The goal is deeper understanding.

Why guardrails are essential

Another major difference between general AI and educational AI is the role of guardrails.

Most conversational AI systems are optimized to keep interactions going, but educational systems require a different objective. Educational AI must know when to continue a conversation and when to stop.

Students need support that remains aligned to specific learning goals rather than wandering into unrelated topics. They also need protection from inaccurate, distracting, or counterproductive interactions.

In many cases, ending a learning interaction at the right moment is just as important as starting one.

Supporting instructors, not replacing them

Sheth repeatedly emphasizes that AI should function as an extension of faculty rather than a replacement for instructors.

Educational AI must align with classroom content, course materials, and teaching approaches. Faculty should maintain control over learning objectives, instructional methods, and the student experience.

The technology can then provide valuable feedback by identifying concepts students struggle to understand and highlighting where additional instruction may be needed.

The future of AI in higher education

Sheth believes higher education’s greatest AI opportunities will come from systems intentionally designed around learning outcomes.

The distinction matters.

General AI helps users find information. Learning-focused AI helps students develop understanding.

As institutions continue evaluating AI strategies, that difference may determine whether technology becomes another digital tool or a meaningful driver of student success.

Lucy streamlines administrative workflow, giving CAPS providers more time to focus on students

The University of Texas at Arlington is piloting an AI tool that helps reduce administrative workload for counselors, giving them more time to focus on students.

Known as Lucy, the tool helps Counseling and Psychological Services (CAPS) staff work more efficiently while preserving the central role of human-centered care.

“Student success is shaped by far more than what happens in the classroom. At UT Arlington, we’re committed to creating an environment where students feel supported, connected and cared for throughout their college experience,” UTA President Jennifer Cowley said. “Innovations like Lucy help strengthen that work by giving our teams more capacity to focus on the people at the center of our mission.”

Lucy, named after the Peanuts character, was designed as a “precision retrieval” tool that provides internal-only information on UTA-specific forms, policies, workflows and documentation guidance, according to Yaroub Saleh, a UTA counseling specialist who created the tool.

“Every minute saved from searching for a form is a minute that can be used to help a student,” Saleh said. “Lucy is a good example of how we can use AI ethically to support our students. And it’s been working. Providers tell me it saves time and gives them accurate information.”

“For example,” he continued, “if a provider is treating a student who is a minor, they used to have to dig through lengthy policy documents that have undergone multiple updates. With Lucy, they can get the exact process to follow and the correct forms in seconds. It’s accurate, consistent and reliable. Because providers receive the same information, it also reduces mistakes.”

Early feedback has been positive, Saleh said, with providers citing simplified day-to-day operations and a reduced administrative burden. Saleh said other universities are already reaching out to explore similar AI tools.

Ultimately, Lucy helps CAPS staff fulfill their mission of helping students increase self- awareness, address mental health and emotional concerns, and make positive changes in their lives. CAPS services are available to all UTA students, with in-person offices in Ransom Hall and the Maverick Activities Center. Virtual care is also offered 24/7 through TimelyCare.

About The University of Texas at Arlington (UTA)

The University of Texas at Arlington is a growing public research university in the heart of Dallas-Fort Worth. With a student body of over 42,700, UTA is the second-largest institution in the University of Texas System, offering more than 180 undergraduate and graduate degree programs. Recognized as a Carnegie R-1 university, UTA stands among the nation’s top 5% of institutions for research activity. UTA and its 300,000 alumni generate an annual economic impact of $28.8 billion for the state. The University has received the Innovation and Economic Prosperity designation from the Association of Public and Land Grant Universities and has earned recognition for its focus on student access and success, considered key drivers to economic

UMGC’s AI strategy starts with governance

University of Maryland Global Campus is approaching AI adoption with a clear institutional principle: innovation only matters if it improves outcomes for students.

President Gregory Fowler describes a strategy built around governance, measurement, and practical implementation rather than experimentation for its own sake. The university has already implemented institution-wide AI training and established an AI Governance Board to ensure adoption remains aligned with institutional mission and student support goals.

The approach reflects a broader shift happening across higher education. Institutions are moving beyond curiosity about AI and focusing on how it can responsibly improve student success and operational effectiveness.

Why UMGC built a closed AI testing environment

UMGC launched nebulaONE as a controlled environment where faculty and staff can safely test AI tools, concepts, and workflows before wider deployment.

More than 300 team members are already using the platform.

The goal is not unrestricted experimentation, it is structured evaluation that allows the institution to identify where AI creates value, where it falls short, and how it can be implemented responsibly.

This type of infrastructure is becoming increasingly important as institutions look for ways to balance innovation with governance and accountability.

How AI is being applied to support students

UMGC is focusing AI adoption on practical student-facing applications.

Conversational AI is helping identify and support struggling learners earlier in the student journey. In the Registrar’s Office, transcript review processes that were previously manual are now partly automated, allowing staff to focus more attention on complex cases that require judgment and intervention.

Career Services has also integrated AI into resume review and mock interview preparation. These tools provide students with more opportunities for practice and faster feedback than traditional one-on-one support models alone can provide at scale.

The focus throughout is operational support that strengthens human-centered services rather than replacing them.

Why measurement matters in AI adoption

AI should function as a strategic enabler, not a replacement for teaching, advising, or institutional judgment.

That requires continuous measurement.

UMGC is evaluating adoption rates, operational outcomes, and areas where systems underperform. The institution then adjusts implementation based on those findings.

This approach reflects a growing expectation across higher education that AI adoption should be tied to measurable student impact rather than broad claims about innovation.

What accountable innovation looks like

The question is no longer whether AI is interesting or technically capable. The question is whether institutions can deploy it ethically, transparently, and in ways that genuinely improve student outcomes.

For UMGC, accountable innovation means governance, human oversight, operational measurement, and a consistent focus on serving learners more effectively.

Jessica Smagler, Head of Research and Outcomes, Kyron Learning

Across higher education, the most common question about AI is no longer “what can it do?” It’s “how do we know it will behave?” That question reflects something important about where the sector stands right now: enthusiasm is no longer the barrier to adoption. Trust is.

Trust in AI isn’t built through promises – it’s built through systems. Without clear internal accountability structures, AI tools operate on good faith alone – and good faith isn’t a governance model.

Institutions evaluating AI-powered tools should look for four interlocking commitments, treated not as product features but as obligations: guardrails, benchmarks, educator control, and a foundation in learning science.

The first line of that accountability is governance – specifically, the guardrails that define how AI is permitted to behave.

Guardrails Built for Learning

AI chatbots routinely welcome off-topic conversations, taking focus away from intended course content and derailing learning goals. Without strict guardrails to ensure that AI behavior stays aligned with education, safety, and institutional expectations, the integrity of the learning experience is at risk.

In the edtech space, guardrails should operate at two levels. One is educational, making sure the AI stays within lesson boundaries, supports reasoning without shortcuts, and redirects students who go off course. The other is around student security and privacy, ensuring student data is protected, sensitive information is automatically redacted, and access to systems remains tightly controlled. And these guardrails should be structural rather than add-ons, built into how the system works from the ground up, not applied as a filter after the fact.

Neither layer is visible to students but both matter to the administrators and instructors who are responsible for what happens in their courses. Together, these are guardrails that institutions can trust and hold companies accountable to, because in education, responsible behavior must be verifiable, not assumed.

Benchmarking for Measurable Results

Guardrails define how an AI system should behave. Benchmarking is how companies prove that it does – and how institutions can hold them to it. Without continuous measurement, guardrails are a promise rather than a practice.

In practice, benchmarking should also operate at two levels. Continuous benchmarking should be run against real learner interactions to detect behavioral drift and measure ongoing alignment with learning objectives. Periodic broader evaluations – run across curated datasets in multiple academic domains – should test for safety, instructional integrity, and consistency.

Critically, institutions should expect AI providers to share benchmarking results openly. A track record that institutions can point to is what separates accountable innovation from well-intentioned experimentation.

End to End Educator Control

AI should amplify great instructors – not replace them. Human oversight is an essential component of AI-powered instruction and should extend across the entire learning cycle, from content creation to the student experience.

At Kyron, for example, no lesson reaches a student without educator review and approval. Instructors set the learning objectives that shape what our AI generates, and retain full editorial control before anything is deployed. This process ensures that what students experience is always aligned to institutional goals.

Educator control should not end once a lesson is deployed to learners. Products should offer visibility into what students are struggling with in aggregate and at the individual learner level, allowing faculty to intervene, adjust, and improve. Insight into misconceptions creates a feedback loop that keeps instructors and institutions informed on student progress.

Grounded in Learning Science

Responsible AI providers don’t stop at governance and oversight. They are also accountable for whether students actually learn. Rooting instruction in established learning science frameworks – like Chi and Wylie’s ICAP model and Vygotsky’s Zone of Proximal Development – isn’t just good teaching. It is a standard that AI providers should hold themselves to and a standard institutions should expect when adopting AI-powered instruction.

Decades of research have made clear that real learning doesn’t happen by giving students answers. It happens when students are encouraged to think critically, reason logically, and develop conceptual understanding. It happens when lessons are intentionally structured to achieve clear learning goals.

A landmark study by Graesser and Person found that 92% of student questions focused on surface facts rather than deeper reasoning, meaning most misunderstandings go undetected and unaddressed. Because students can so easily appear engaged without building true conceptual understanding, instruction must be intentionally designed to surface reasoning and guide deeper thinking.

When AI providers root their tools in learning science, they are making a verifiable commitment to student outcomes – not just a promise of engagement.

This is Accountable Innovation

AI in education demands more than innovation, it demands accountability. Institutions have a right to ask AI providers how they know their tools will behave, and AI providers have an obligation to answer concretely. Organizations using AI must do so in ways that are safe for students and transparent to institutions. By setting guardrails and benchmarks, keeping educators in control throughout, and grounding tools in learning science, we can be confident that we are innovating responsibly.

At Kyron, our commitment to safe, accountable AI has helped us build enduring partnerships with institutions like Miami Dade College and Western Governors University, while creating opportunities to collaborate with forward-looking colleges like Rio Salado College and established curriculum companies like McGraw Hill. When we answer questions with evidence rather than promises, we build the kind of trust that makes responsible AI adoption possible, both for our partners and for the students they serve.

Interested in learning more about Kyron Learning? Visit www.kyronlearning.com or connect with our team to get started.

What does accountable innovation mean in higher education

Accountable innovation is becoming the defining expectation for higher education leaders.

Across the Presidents Forum network, this means designing new models that are not only innovative but also measurable. Institutions are focusing on flexible pathways for working learners, stronger alignment between education and workforce opportunity, and delivery models that reflect how students actually live and learn.

The emphasis is clear: innovation must lead to better outcomes, not just new ideas.

How institutions are putting innovation into practice

Institutions are translating this principle into concrete changes.

Flexible scheduling and online delivery are being paired with clearer pathways to completion. Programs are being designed with employer input to ensure relevance. Student support models are expanding to address barriers outside the classroom.

These changes reflect a broader shift. The goal is not access alone. It is access that leads to completion, employment, and long-term mobility.

What is happening in federal policy right now

At the federal level, the Department of Education’s AIM negotiated rulemaking is reinforcing this shift.

The first week of negotiations signaled a move away from a compliance-driven system toward one focused on outcomes, value, and consumer protection. At the same time, the proposals introduce new expectations around transparency and legal compliance.

Negotiators worked through a large portion of the draft text, but key issues remain unresolved. Areas such as outcomes-based accountability and accreditor flexibility continue to generate debate.

The second round of negotiations, scheduled for May 18 to 22, will be critical in shaping how these policies take final form.

Why policymakers are focused on outcomes and value

The direction of policy reflects broader public expectations.

Students, families, and policymakers are asking more direct questions about return on investment. They want to understand how education leads to employment, earnings, and career advancement.

This is driving a shift toward program-level outcomes, clearer disclosures, and stronger accountability frameworks. Institutions that can demonstrate value will be better positioned in this environment.

What this means for institutional strategy

Higher education is entering a period where innovation and accountability are closely linked.

Institutions will need to align program design, student support, and data systems with clear outcome measures. They will also need to communicate those outcomes effectively to policymakers and the public.

The opportunity is significant. Institutions that can demonstrate both innovation and results will define the next era of higher education.

The bottom line

Accountable innovation is no longer optional.

It is the standard by which institutions will be evaluated, funded, and trusted.

By Anne M. Kress, President, Northern Virginia Community College

Higher education leaders and policymakers could be forgiven for making AI the center of every conversation about preparing students for a world of work changing at record speed. The numbers are striking: a Lightcast study found that one-third of the skills required for the average job changed between 2021 and 2024. A LinkedIn executive observed in a May 2025 New York Times opinion piece that AI was breaking the bottom rung of the career ladder — the entry-level roles where generations of young workers got their start—and that was a year ago.

AI deserves our attention. But it isn’t the only issue that does.

At Northern Virginia Community College (NOVA), we hear consistently from students and employers that career-connected learning is often the difference between a graduate who gets hired and one who doesn’t. Students pursuing IT and engineering at NOVA already benefit from that connection: internships and apprenticeships with partners like Micron, Digital Realty, Microsoft, and AWS, plus the opportunity to earn employer-valued credentials on the path to their certificates and degrees. These opportunities build careers.

They also build the durable skills that employers – not just in STEM fields – say they need in their early-career workers. A study last year by Presidents Forum member Western Governor’s University and UpSkill America defined durable skills as the “enduring skills that are not job/role specific but are valued across all roles and workplaces (teamwork/collaboration, active listening, communication, etc.).” It also noted a prevailing belief among employers that skills needed to succeed on Day One of a job (trustworthiness, attention to detail, collaboration, integrity) are critical – and gained through real-world experience rather than academic instruction.

How do we equip students, regardless of discipline, with the durable skills needed to thrive in today’s whirlwind workplace?

In fall 2022, more than 44,000 students were enrolled in liberal arts courses across nearly 2,000 sections at NOVA. With support from the Jack, Joseph and Morton Mandel Foundation and partners in the business community, NOVA launched an initiative to expand our career-readiness infrastructure to these students. The result is a two-part model: a micro-credential program that makes the skills embedded in a humanities education visible and verifiable to employers, and a micro-internship program that puts those skills to work in real professional settings.

We started, as we always do, by asking employers what they actually need and value. We convened a group of 40 professionals working in humanities-adjacent fields and posed a direct question: What does an emerging professional need to succeed? Over eight weeks of structured discussion, faculty synthesized employers’ responses into five defining characteristics: workplace humility, adaptability, a willingness to learn, strong communication, and technical fluency. Those conversations became the foundation of a micro-credential program comprising 24 digital badges across three pathways — Critical Thinking, Communication Skills, and Leadership. These are employer-informed markers of specific, validated competencies — designed from the outset for college-wide adoption, so any NOVA student, regardless of discipline, can build and demonstrate these skills.

The micro-internship program grounds those credentials in real experience. These are short-term, project-based, remote or hybrid engagements that fit the realities of students’ schedules: over 70% of NOVA students are part-time, juggling classes, jobs, and caregiving, so we wanted the micro-internships to be accessible and achievable. We also wanted the students’ work to be consequential. NOVA students helped Smithsonian curators sort through a newly acquired collection of 19th century postcards. Others analyzed truancy data for a Chatham County judge, created content celebrating the Alexandria Film Festival’s 20th anniversary, and documented campus life for NOVA’s marketing office. These micro-internships are not simulations. They are real projects, for real organizations — exactly the experiences that help a student walk into a job interview and say, with evidence, what they can do.

The skills these students develop — synthesis, communication, ethical reasoning, adaptability — are hardest to automate and most valuable in a world reshaped by AI. Thanks to funding from the Mandel Foundation, NOVA has been able to build the infrastructure that supports the development and demonstration of these skills. Through this project we have learned that our students are ready and their prospective employers are willing. The only question is whether more higher education institutions can follow NOVA’s lead and meet them both halfway.