Meet Cog: One AI Assistant on Every Page of Your Academy

There is a question every academy asks in the second week of using an AI learning platform, and it is never the one the demo prepared you for. It isn't how good is the AI? It's where is it? — asked by a trainer who found the tutor inside the lesson player on Monday and cannot find it from the course editor on Tuesday.

That question is a design failure, and it was ours as much as anyone's. This is the story of how we answered it: Cog, the EduGears AI assistant surface — one assistant, one place, every page.

Why were there four chats in the first place?

Because each one was built where it was needed, and each one was right on its own. A tutor belongs beside the lesson a learner is reading. An assistant belongs beside the resource an instructor is editing. A full-page chat belongs where you go to think in long form. Every surface was a sensible local decision, and together they added up to something nobody would have designed on purpose.

The deeper cost was in the conversations themselves. Four surfaces meant four threads, each with its own memory of what you had already explained, and moving between screens meant starting again. The AI knew your curriculum. It did not know your morning.

What is Cog?

Cog is the single assistant surface for EduGears AI LMS. Once you are signed in, a round Ask Cog button sits in the bottom-right corner of every page. Click it and a panel opens over the page — a floating card on a computer, the whole screen on a phone. That is the entire interface. There is no second place to look.

• It is on every page. Dashboard, course editor, gradebook, calendar, a lesson, a quiz result — the bubble is in the same corner on all of them.
• It survives navigation. Walking from a lesson to the calendar and back neither closes the conversation nor restarts it. The panel simply learns where you now are.
• It is yours to size. On a computer, drag its left edge for width, its top edge for height, or the corner for both; double-click an edge to put that dimension back. The size you settle on is remembered in that browser.
• It keeps its history. Close it with the × and reopen it later and you pick up where you left off, with earlier threads still reachable.
• It costs nothing until you use it. Nothing loads until the first time you open it, and each request shows up in Reports → AI usage like every other AI feature — one row per request, including whether it ran on your own API key or on the platform's built-in AI.

One thing Cog is not: a new AI tool. The tutor and the assistant you already had are the same tutor and assistant, with the same grounding and the same 26-tool suite behind them. Cog is where you meet them now.

How does it know what I'm looking at?

Because the page tells it, before you type anything. On a course or a lesson, Cog is given the title of what is in front of you; on any other page it is given the help topic that page belongs to. So "how do I do this?" is answered about the screen you are on rather than about the platform in the abstract, and "explain this" means this.

What it is currently grounded in is not a guess you have to trust — it is printed in the header as a context chip: Resource, Lesson, Module, Course or General. Click the chip and you can widen the scope: from the resource up to the lesson, the module, the course, or right out to General — anything, no page context. Narrowing is never offered, because narrowing is what navigation is for.

• It follows you. Move around inside a course and the chip updates while the conversation continues.
• It re-grounds deliberately. Open a different course and Cog re-grounds there and starts that course's own thread — a conversation grounded in the course you just left is worse than a fresh one.
• It has a sensible answer off-course. On your dashboard it stays on whichever course you last opened it from, or the first course in your list if you haven't opened one yet. With no courses at all it says so, instead of opening an empty box.

This is the same grounding argument we made for the tutor itself — that context is what separates a teaching tool from a chat window with manners. If you want that case in full, it is in AI Tutors That Know Your Course. Cog is that argument applied to the interface rather than the model: the assistant should not need to be told where it is standing.

Do learners and instructors get the same thing?

No — and the difference is enforced by the platform, not by the panel. Learners get the AI Tutor, which explains and coaches on the course material. Instructors and admins get the AI Assistant for authoring and teaching help, and can switch to the Tutor whenever they want. A learner cannot reach the assistant by any route, whatever the panel is showing, because the permission is checked where the answer is produced.

Inside the Tutor, the teaching styles came along unchanged: Tutor for step-by-step explanation, Mentor for study strategy, Socratic for discovery by question, and Coach for exam practice — plus learning modes (Adaptive, Clear Doubts, Exam Prep, Concept Map) for the shape of the session. Switching style starts a fresh conversation, because a transcript belongs to the style it was held in.

Can an instructor see what the learner sees?

Yes, and this is the part instructors like most. When an instructor toggles from the Assistant to the Tutor, they are not getting a demo mode or an approximation — they are getting the learner's Cog exactly as a learner would get it, with the same answer-stripping rules applied to the same course material.

That turns a policy question into an observation. "What will it tell my students about the case study before the assessment?" stops being something you reason about and becomes something you check, on the actual resource, in about fifteen seconds — with the answer in front of you rather than in a support ticket.

Is it safe to leave on during assessment?

Yes. Before and during an attempt, Cog will not surface quiz or answer-key material: a learner sitting a quiz gets hints and explanation of the underlying concept, never the answer to the item in front of them. Full explanations unlock after a graded attempt, which is exactly when they are worth the most — the learner has committed to an answer and is finally curious about the right one.

This matters more for what it prevents than for what it enables. The alternative — an assistant that has to be switched off for every assessment window — is an administrative burden that gets forgotten once, and once is enough. Cog is built so the safe configuration is the default one.

Everything else about assessment integrity works as it did: lesson plans and answer keys stay out of the learner-facing player, marking still flows into the gradebook, and the one weighted course grade every report agrees with is unchanged.

Does it know how the platform itself works?

It does, and that is the quietly useful half. Cog is grounded twice over: in your course's own materials through Curriculum Intelligence, and in the platform's own help documentation. So "how do I move this activity into another section?" is answered out of the documentation for the page you are standing on — not out of a model's memory of some other product's menus.

When the Assistant answers a question of that kind, it cites the help article it drew on, so a staff member can read the source and learn the platform rather than re-asking it every week. The Tutor deliberately does not cite: a learner asking about photosynthesis does not need a footnote to an admin manual.

For an academy this replaces a familiar bottleneck. The person who knows the platform best is usually the busiest person in the building, and every "quick question" lands on them. A grounded assistant on the page where the question occurs is a better first answer than a colleague's memory, and it is available at eleven at night.

What do we have to do to switch it on?

Nothing. Cog is on for every tenant, existing and new. There is no flag to request, no plan to upgrade to, no migration of past conversations to schedule. If you are already running EduGears AI LMS, open any page and the bubble is in the corner.

The older routes into the same AI — the course editor's AI Tutor and AI Assistant entries under Course Tools, and the course view's AI Tutor button — are gone, replaced by the one dock. That is the trade being made deliberately: one fewer thing to teach a new trainer, and one fewer place for a learner to fail to find help.

The same Cog runs in the other EduGears AI products too — inside Moodle, Canvas and other LMSes through the LTI tool, and in the Studio publisher workspace — so an institution running more than one of them teaches its staff one interface, not three.

FAQ