How will agentic AI show up in higher education?
While higher education continues to come to terms with generative AI, a new buzzword is beginning to capture imaginations: agentic AI. Gartner named agentic AI the top tech trend of 2025 and forecasts that by 2028, 33% of enterprise software applications will include it. But what is agentic AI? When and how will agentic AI show up on campus?
In this blog post, I’ll draw from my work on EAB’s IT Strategy Advisory Services team to define agentic AI and explore how it will reshape AI applications in higher education.
Defining agentic AI
-
What is agenic AI?
Agentic AI can reason, make decisions, execute tasks, and navigate different environments independently, without constant human guidance.
Agentic AI systems use AI agents (often teams using generative AI) to take proactive action within predefined limits, without requiring human prompting at every step. They can iterate and experiment to achieve goals and tasks, even in unclear scenarios. With knowledge and memory recall, they can learn and improve from past experiences. These systems also regularly combine multiple AI models, optimizing them for certain tasks.
Agentic AI systems use AI agents (often teams using generative AI) to take proactive action within predefined limits, without requiring human prompting at every step. They can iterate and experiment to achieve goals and tasks, even in unclear scenarios. With knowledge and memory recall, they can learn and improve from past experiences. These systems also regularly combine multiple AI models, optimizing them for certain tasks.
Agentic AI builds on generative AI, often leveraging teams of agents using generative AI to execute more complex tasks across administration and academia. Almost three years since generative AI emerged, most applications are still a far cry from the autonomous coworker or across-the-lifecycle tutor initially touted. Today, higher education is most likely to have deployed AI applications that retrieve information to support small-scale decision-making (e.g., an HR policy chatbot). Agentic AI not only optimizes existing generative AI use cases but also expands the scope of what they can achieve. This progression is already evident in applications out-of-sector. Early coding assistants like Github Copilot generated code and enhanced developer productivity, but they did not think through application development. With the advent of agentic AI, Cognition Labs recently deployed the software engineer assistant, Devin, who can write code as well as debug and deploy entire applications autonomously
In higher ed, agentic AI can enable AI tutors that can not only recall course information but also troubleshoot learning challenges more effectively and reach out to advisors on a student’s behalf.
Here are 3 examples of generative AI applications are already evolving on campus due to agentic AI:
-
Academics
GenAI use case
Students ask a generative AI-powered TA questions about uploaded course materials. The TA tool provides evidence-based responses but does not elaborate or coach students through material.
Agentic AI advancement
University of Michigan recently launched a virtual teaching assistant (TA) pilot for over 9,000 business students, driven by agentic AI from Google Gemini. The virtual TA walks students through course concepts, guiding them through problems and emphasizing their learning without giving answers away. At the same time, faculty receive analytics on student interactions and performance.
-
Administration
GenAI use case
A generative AI chatbot answers prospective student inquiries like what an academic program’s degree requirements are.
Agentic AI advancement
Ithaca College is currently developing an agentic AI system called Aurora, which is designed to actively guide students through a wide range of processes. This system aims to streamline tasks that currently require students to navigate multiple systems manually (e.g., selecting electives). In addition to directly answering questions with student data, it will be able to support complex administrative processes, alert students to important deadlines, and simplify interactions with various campus systems (e.g., proposing courses based on a student’s degree requirements). Aurora’s proactive approach is intended to reduce complexity and help students stay organized and focused and leverage college services.
-
Research
GenAI use case
Given a research prompt, GPT-4o examines available resources (e.g., survey responses, journal articles) and generates a research outline, filtered largely through a single lens unless prompted to consider discipline-specific approaches.
Agentic AI advancement
Stanford University developed the Virtual Lab to simulate research teams with the goal of facilitating interdisciplinary research. A large language model (LLM) principal investigator agent orchestrates a team of LLM agents with assorted scientific backgrounds to examine interdisciplinary approaches to research topics.
How campus stakeholders will begin engaging with agentic AI
Faculty, staff, and students will encounter and adopt agentic AI in distinct ways. Here’s how we think agentic AI will begin impacting these different groups:
1. Faculty and staff will access agentic AI at scale through enterprise software applications (e.g., M365 Copilot)
Mainstay vendors like Microsoft are already beginning to integrate agentic AI capabilities within their products. In March 2025, it announced two new deep-reasoning agents embedded in Microsoft 365 Copilot. Microsoft’s Research and Analyst agents can access work data from platforms like Salesforce and ServiceNow to support in-depth research and reporting. For example, the Analyst agent is positioned as an expert data scientist, able to analyze raw data, write Python code, and generate full reports like revenue projections autonomously.
See more examples of how more enterprise applications are incorporating agentic AI:
| Vendor | Agenic AI feature | Sample use case |
|---|---|---|
| Workday | Role-based AI agents in areas such as payroll, financial auditing, and policy that can be configured to support role-specific tasks. | Role-based AI agents in areas such as payroll, financial auditing, and policy that can be configured to support role-specific tasks. |
| Box | AI agents designed to streamline document navigation include a search, deep research, enhanced data extraction, and Microsoft 365 Copilot and chat agent. | A legal team could use the Box AI agents to examine indemnification clauses across enterprise contracts, report on trends and outliers, automatically classify contracts, and update file metadata accordingly. |
| Navigate360 | AI agents across the student lifecycle, from course planning, reporting, to marketing campaign agents. |
Enrollment marketing staff can use campaign agents to analyze historical data, identify prospective students with a high likelihood of enrolling, and create hyper-personalized communications directed at them. Once enrolled, faculty and staff will be able to anticipate prospective student concerns based on their profile history and monitor ongoing holistic performance with the student insights agent. |
| Amazon | Amazon Q is an agentic AI assistant that is being rolled out across Amazon products. |
In Amazon Q Business, users can use Q to build apps to automate workflows and tasks like drafting emails from notes and updating associated records. In Amazon Connect, the cloud contact center, Q can detect customer issues, respond automatically, and provide recommended actions. |
2. Administrative staff will automate repetitive processes (e.g., procurement and IT ticket routing) because agentic AI reduces implementation barriers
While agentic AI significantly expands what can be automated effectively, it is still best suited for structured, well-defined workflows. Compared to earlier automation technologies like machine learning and robotic process automation, deploying agentic AI typically requires less manual configuration and oversight for similar use cases. And as vendors rush to integrate agentic AI, conditions are ripe for scaling automation. For higher education staff, this offers the opportunity to offload rote work and decision-making, leaving more time for them to focus on higher-order decision-making.
Vendor solutions are already emerging. In IT, ServiceNow agents can now detect server problems, propose resolutions, execute fixes, and even review results while keeping staff informed throughout the process. Similarly, ServiceNow’s HR agents can guide new staff through onboarding, identifying what they need, and automatically provisioning access after cross-referencing company policies. During any step in the process, agentic AI systems can directly raise issues with different stakeholders.
3. Students will incorporate agentic AI tools that traverse browsers and computer files into how they undertake schoolwork
Enterprising students will soon discover how much more they can accomplish with browser- and computer-using agents that can interact with the internet and their computer files and programs to achieve complex tasks. Manus AI is the latest highly publicized multi-agent system from China with free deep research capabilities. With Manus, a central taskmaster oversees different agents completing subsets of tasks and responsibilities, whether one is analyzing data or developing code. Domestically, OpenAI recently previewed its own research agent called Operator.
While a student might now prompt ChatGPT to draft their paper, they will soon be able to ask agentic AI tools to draft a research plan, test hypotheses, iterate on findings, and generate a research report—all with one prompt. The good news is that faculty have already begun designing AI-proof and AI-first assignments, but the urgency to do so has intensified with the emergence of agentic AI.
-
Help faculty use AI effectively
Access our Construction Kit for Faculty AI Resources toolkit to help engage faculty with AI in their curricular and teaching practice.
In Conclusion
Amid all the hype, it can feel overwhelming to try to keep up with AI advancements lately. But agentic AI is worth digging into; its potential to power automations that take over rote work could transform the day-to-day experience of staff across your institution.
Whether you’re an AI front-runner or just beginning to develop your school’s approach to generative AI, EAB is here to help, and I’ll continue to share what I’m learning from colleges and universities embracing the transformative power of this fast-changing technology.
More Blogs
How AI-powered ad filters are disrupting enrollment marketing campaigns
Where superintendents are creating clarity in 2025
