Designing an AI companion that understands where you are, where you're going, and what might help you get there.

MyGooru already held years of knowledge about its learners: their goals, beliefs, progress, competencies and personalized pathways.

But the way people used AI was changing. ChatGPT and Claude had made conversation a familiar way to work with it. That raised a different question for us:

What if users didn't have to navigate the system to find what they needed? What if the system could understand their context and meet them there?

That became the Companion. It couldn't be one more chatbot.

A learner asking “What should I learn next?” needs something very different from an instructor asking “Who in my class needs help?” So the job became designing one Companion for different people, contexts and goals.

Learner: help me move forward. Start learning from almost anywhere. The Companion understands the learner through their context, beliefs, goals, progress and Portrait of a Graduate, and connects them to pathways, nudges and next steps that fit.

Instructor: help me understand my learners. See where learners are and who may need support, get recommendations, and assign activities that could help them move forward.

Mentor: help me guide my mentees. Understand a mentee's progress, learning journey and pathways without piecing their context together from scratch.

The learner's home opens on a question, not a menu: “What would you like to learn today?” They can type or speak something as ordinary as needing help with fractions before Friday's quiz.

Under it sits only what helps them keep going: what is due this week, time spent, their streak, and a card that picks up where they left off.

For a learner, the Companion looks inward: where am I, and what should I do next?

Pick up where you left off

Classify Polynomials by Degree

Here you will learn how the highest power names the shape.

Continue · 14 min →

This week

2 due · 4h 20m · 12-day streak

For an instructor, the Companion looks across learners: where is everyone, who needs attention, and what can I do?

On the Learning Journey it reads the class milestone by milestone, and says why the class is held up, not only that it is.

Why the class is held up

They all stopped at the same lesson, and it is not the hardest one.

22 of 24 cleared the lesson before. 6 cleared this one — and 5 of those worked on paper first.

Assign the paper-first version →

In Class Activities it reads what was assigned and points at the one activity that needs the teacher.

The one that needs you

The slide deck is not a marking problem.

Average time was three minutes on a fifteen-minute deck. They clicked through it.

Show me what they skipped →

And when work comes in, it puts the grading in order: what it can mark itself, and the written answers only a teacher should.

The order I would take them in

The five written pieces carry all the information.

Fyodor and Gabriel made the same wrong turn — one reply covers both.

Open the five written →

A mentor is linked to a learner, never assigned one. The learner approves the request, sees exactly what the mentor will get (dashboard, learning journey and proficiency, all read-only, plus a weekly summary) and can revoke it from their profile.

For a mentor, the Companion follows the individual over time: how is my mentee progressing, and where might I support them?

Actionable insight

Anna needs a nudge

Stuck on the arithmetic quiz for 4 days — a nudge unblocks the next section.

Actionable insight

1 assignment overdue

Across your learners this week — worth a quick pass tonight.

Actionable insight

Maria is on a 4-day streak

Worth a “well done” tonight.

From there the mentor can open the learner's own views: dashboard, learning journey and proficiency. They don't have to rebuild the context themselves.

The Companion can hold a conversation, but most of the product sits behind the conversation.

As someone uses MyGooru, the system builds up a picture of who they are, what they believe, where they are, where they're going and how they might get there.

So a conversation doesn't start from zero. What the system knows carries into what happens next.

The Companion doesn't simply generate the next piece of content from a prompt.

Behind it, EcPA rules help decide what content comes next, based on the learner and how their context has changed. So a pathway isn't a fixed sequence.

For the learner it can feel simple. Underneath, the system keeps deciding what is relevant next.

Because the Companion carries context, learning doesn't have to begin from a traditional course page.

A learner can enter from a question, an interest, a pathway, an activity or something happening in their current context. The Companion connects that moment back to their learning journey.

That moved the design from navigation built around where content lives to an experience built around where the learner currently is.

Start anywhere

→

Understand context

→

Connect to pathway

→

Continue learning

For a learner, the Companion looks inward: Where am I? What should I explore next?

For an instructor, it looks across learners: Where is everyone? Who needs my attention? What can I do?

For a mentor, it follows the individual over time: How is my mentee progressing? Where might I support them?

The learning system underneath is shared. Only the perspective changes.

The work started with Learner, Instructor and Mentor, but the idea underneath wasn't limited to education. It could apply in very different settings.

That mattered later, when we began exploring the Companion in other domains.

But Learning was where we first had to make it understandable and useful to the person using it.

Working on the Companion changed how I think about AI products.

The chat interface was the smaller part. The real work was making an intelligent system understandable to the person using it, while most of the rules and reasoning stay underneath.

The interface became simpler. The system behind it became more contextual.

These are the screens as they stand now. Each role has its own look, so the three never blur into one product with three skins.

For the learner, the home opens on a question, not a menu. The journey sits on one path: milestones, competencies and the next resource on a single line, so it’s always clear where you are.

For the instructor, the Companion sits beside every page and never covers it. It sorts what needs the teacher by how soon it matters, and reads pace from today, so the class that won’t finish shows up weeks before the deadline.

Before drawing any of it, I mapped where the Companion lives on each screen and what it is allowed to talk about there. On Home it reads all six classrooms; on one student it reads only them. Every nudge opens and closes the same way, so sixteen kinds of nudge are one thing to learn.

Grading is where that rail has to work hardest. The rubric is drafted if there isn’t one, short answers are pre-marked, and the written ones come one at a time with a suggested mark the teacher can change.

For a mentor or parent, the question is who needs a word tonight. The home lists each learner with what’s due and what’s stuck, and opens straight into their journey.