EdTech Opportunities Case Study
Agentic AI & Automation Solutions for Educational Institutions
Prepared by Umer | Agentic AI Engineer Intern
Overview
This document outlines six agentic AI opportunities for educational institutions, developed as part of the
internship case study assignment. Admissions and inquiry handling have been intentionally excluded, as
the company is already working in that area. The AI-in-education market is projected to reach roughly
$32.27 billion by 2030, and by the end of 2026 an estimated 40% of enterprise applications are expected
to embed task-specific AI agents. The ideas below focus on where agentic AI, meaning systems that
reason through multi-step tasks and act autonomously rather than simply responding to prompts, can
meaningfully reduce manual workload for students, faculty, and administrators.
1. Multi-Agent System for Students
Rather than building a single do-everything assistant, this idea proposes a multi-agent architecture where
specialized agents handle distinct responsibilities and an orchestrator agent routes each request to the
right one, combining answers when a query spans more than one domain. This structure is also what ties
the rest of this document together, since several of the agents described in later sections (grading,
timetable, LMS, attendance) can plug into this same orchestrator rather than existing as separate
disconnected tools.
Core sub-agents:
• Fee/Finance Agent - handles fee status checks, payment reminders, and document requests without
manual front-desk involvement.
• Academic Agent - answers curriculum and course-progress questions.
• Scheduling Agent - connects to the timetable/rescheduling system described in Section 3.
• Wellbeing Agent - surfaces the engagement and academic-help signals described in Sections 5 and
6.
• Orchestrator Agent - routes each student query to the correct specialist and merges responses for
multi-part questions.
2. CV-Based Monitoring & Proctoring System
A computer-vision based system to support exam integrity and general campus monitoring, covering both
students and faculty. During exams, the system flags irregular behavior such as repeated looking away,
unusual device presence, or multiple faces in frame, producing a log of flagged clips for a human
invigilator to review rather than issuing automatic penalties. Outside exams, the same pipeline can extend
to broader irregular-activity detection, such as unattended lab equipment, access to restricted areas, or a
scheduled class not actually taking place in its assigned room, which also feeds useful data back into the
timetable system in Section 3.
Because this system involves monitoring people, including faculty, it should always retain a human-
review step before any consequence is applied, and its scope and policy should be communicated
transparently to everyone covered by it.
3. Timetable & Makeup-Class Rescheduling Agent
This idea comes directly from a real problem observed by me at my own university: timetables are
maintained manually by a single person, which regularly causes scheduling clashes as departments,
rooms, and faculty availability change. An agent-based system would hold all scheduling constraints
(teacher availability, room capacity and type, department rules) and reason through trade-offs when
building or adjusting the timetable, rather than relying on one person's manual tracking.
Key capabilities:
• Automatically detects and resolves scheduling clashes across departments, rooms, and faculty.
• Handles makeup classes end-to-end: when a teacher misses a class or requests a reschedule, the
agent finds valid replacement slots, proposes options, and notifies students and faculty once
approved.
• Manages repeating students separately: if a student is retaking a course, the agent checks that the
retake slot does not clash with the student's regular current-semester schedule before assigning it.
• Explains its own decisions, e.g. stating exactly which constraint is blocking a slot ("no lab room
available before Friday") instead of failing silently.
4. Grading & Checking Agent
This agent automates grading for objective assessments (multiple-choice, numeric answers, code test
cases) end-to-end, and assists with subjective grading (essays, short answers, code style) by scoring
against a rubric and generating personalized feedback for the student. Borderline scores or suspected
plagiarism/AI-generated submissions are flagged for teacher review rather than finalized automatically,
and higher-weight assessments such as final exams should keep a human confirmation step before results
are released.
5. Enhanced LMS Agent
Beyond a basic FAQ chatbot, this agent answers course-content questions using the institution's own
LMS materials, explains grades and feedback produced by the grading agent in Section 4, and can take
actions on a student's behalf, such as setting deadline reminders or submitting a resubmission request,
which is what makes it agentic rather than a simple search interface. Two additional features are included
as part of this agent rather than as standalone systems:
• Academic-help meeting scheduling: when a student is struggling, the agent identifies the relevant
professor and books a meeting directly, pulling the professor's free slots from the
timetable/rescheduling agent in Section 3.
• Curriculum-gap suggestions: the agent compares a student's performance against the curriculum and
relevant industry trends, and suggests electives or resources to help close identified gaps, delivered
within the same chat interface.
6. Attendance & Engagement Monitoring Agent
This agent tracks attendance, LMS logins, watch-time on lecture content, and quiz participation to
identify students who are falling behind. Its key value is being proactive: it notifies a student before a
problem becomes serious, for example warning them as they approach an attendance cutoff rather than
only after they have failed the requirement.
This agent is connected to the LMS agent in Section 5: rather than operating as a separate dashboard, its
notifications, along with the meeting-booking and curriculum-gap features, are all delivered through the
LMS agent's interface. The result is that a student experiences a single assistant, even though several
specialized agents are working behind the scenes to support them.
Closing Note
Together, these six ideas form a connected ecosystem rather than six unrelated tools: the multi-agent
orchestrator in Section 1 ties into the timetable system in Section 3 and the LMS agent in Section 5, while
the CV-based monitoring system in Section 2 and the grading agent in Section 4 each feed relevant
signals back into that same ecosystem. This structure is intended to demonstrate practical, deployable use
cases for agentic AI in an educational setting, with reasonable human-oversight checkpoints built in
throughout.