Most people have met a chatbot. An agent is different: it can take steps — open your own course reports, read what is actually in them, and only then answer. Here is what that means, and what it looks like inside the Moodle, Canvas, Blackboard or Brightspace course you already teach in.
No jargon. Three short answers.
A plain chatbot answers from whatever it already knows, in one pass. An agent behaves more like a colleague with your course open beside them: it works out what it needs, opens the right report for that course, reads the rows that are actually in it, and only then writes an answer. The numbers come from your class, not from the model's memory.
Agentic simply means the AI can do something — run a lookup, use a tool, go back for one more detail — instead of only producing text. The limits matter as much as the ability: which tools it may use, whose data it may open, what it may do with what it finds. In EduGears AI the only tools the agent has are read-only lookups into the same reports your LTI 1.3 integration already shows you.
Example
How did the class do on the Module 2 quiz?
Eighteen of the 22 learners in Biology 101 have submitted the Module 2 quiz, and the average score is 64%. Eleven scored above 70% and seven below 50%. Four haven't attempted it yet: Amara Osei, Daniel Ruiz, Priya Nair and Tom Becker. Want me to check how those four are doing on the module itself?
Where this came from
Quiz Submissions · 18 rowsCog opened the same Quiz Submissions report you can open yourself, counted the rows and linked it. Click the source and that report opens in the EduGears AI tool inside your course, with the filters already applied — so every figure it just quoted can be checked in one click.
Cog is the assistant that already sits on every EduGears AI page. Answering questions about your class is something it gained — nothing new to install, and nothing for your LMS administrator to register again.
Instructors stay in Moodle, Canvas, Blackboard or Brightspace. They open the EduGears AI activity in their course, click Ask Cog in the corner (or press Cmd/Ctrl + K) and the chat opens beside the work, with the LMS identity they launched with. No second login, no separate analytics app, no export.
Module progress, at-risk learners, quiz submissions, score distribution, assessment submissions and your course roster — the same reports the integration already shows you. Ask “who hasn't started Module 3?”, “how is Module 2 going?” or “who is falling behind?” and Cog picks the right report, applies the filters and reads the rows.
Under the answer sits a chip naming the report it used and how many rows it read. Click it and that report opens with the same filters applied, so a figure you want to double-check is one click from the table it came from.
Cog carries two personas. Instructors open on the AI Assistant, and that is where class-data questions are answered. Learners only ever get the AI Tutor, grounded in their own lesson and assessment-safe; the data side is instructor-only and enforced on the server, not merely hidden in the interface.
The limits are built into how the agent runs — not merely written into its instructions.
Cog can look; it cannot change anything. No grade touched, no submission edited, no message sent, nothing posted to your gradebook. Reading is the only thing it is able to do: its tools have no way to write, and the database connections behind them are read-only as a backstop.
The agent runs as the person asking, on the platform they launched from: same role, same courses, same permissions your LMS already grants over LTI 1.3. A course an instructor cannot open is not open to Cog while they are the one asking.
Answers name the report they came from and link to it with the filters applied. Nothing is recalled from memory: where a report has no rows for the question, the answer says so rather than filling the gap.
Class-data questions are a feature your LMS administrator can switch on or off for the whole platform, at any time, from the Features page in the EduGears AI management portal.
An AI agent is an AI that takes steps to answer you instead of replying from memory alone. It works out what it needs, uses a tool to look it up — in our case one of the reports for your own course — reads the result, and writes an answer based on what it actually found.
Agentic AI is AI that can act — use a tool, run a lookup, go back for one more detail — within limits somebody sets. The limits matter as much as the ability. In EduGears AI the agent's only tools are read-only lookups into reports the person asking can already open.
Questions your reports can answer: who hasn't started a module, how far a class has got, who is falling behind, how many learners submitted a quiz and how they scored, who is on the course roster. Cog picks the report, applies the filters and answers from the rows it read.
No. Cog is read-only by design and by construction: it has no tool that writes, edits, sends or deletes, and nothing it does reaches your LMS gradebook. Marking stays a human action in the normal screens, where your approval is still what posts a score back over LTI 1.3 AGS.
Students get the AI Tutor, grounded in their own lesson and assessment-safe — it never reveals quiz questions or an answer key before their attempt is graded. Class-data questions are instructor-only, enforced on the server.
No. It is part of the EduGears AI LTI 1.3 integration your administrator already registered — no plugin, no second registration, nothing for instructors to set up. Administrators can switch class-data questions on or off for the platform, and AI is included on paid plans, subject to fair use.
Get started in under 3 minutes. No credit card required. No intrusive plugins. Just connect and go.