The Tutor That Doesn't Leave When Class Ends

ABI is an artificial intelligence platform designed to support students beyond the classroom: answering questions, guiding study paths, and tracking individual progress. Its greatest benefit proved to be human rather than technological: ensuring no student leaves with their questions unasked.

Anyone who has ever taught a class knows the scene. The bell rings, and while packing up, two or three students approach the desk to ask questions. It’s always the same ones: the confident ones, those who don't have to rush off to work, the ones who arrived with their questions already formulated. The rest leave with their questions bottled up inside. Sometimes they figure it out that night. Most of the time, they abandon it altogether.

That scene was the starting point for ABI, an AI assistant I developed to extend support beyond the classroom. The question that sparked it wasn't technical: How do we keep course content accessible after class has ended?

What It Does, in Plain English

ABI isn't a search engine or one of those chatbots that answer just anything. It’s closer to having a tutor available 24/7 who knows the lectures, exercises, course notes, and official documentation inside out.

When a student asks something, ABI searches that material for the exact passages that address the question, crafts an explanation, and—here's the key point—cites where every piece of information came from. The student doesn't get an answer out of thin air: they get an answer along with a clear path back to the original source. An assistant that answers without citing sources asks you to trust it. One that answers with sources invites you to verify. In education, the latter is the mindset we want to build.

What I Didn't Expect: Answering Wasn't the Problem

When I showed early versions to students, I expected them to value the answers. And yes, they did. But what they appreciated most was something else, articulated better by one student than I ever could have written:

"The agent helps by guiding me through the study path—it doesn't just wait for my questions, it maps out the trail. That saves me when I'm lost or when I don't have anything to ask."

"I don't have anything to ask." That comment forced me to rethink the entire project. I had assumed students arrive with a clear question and just need an answer. In many cases, the opposite is true: students know they didn't understand something, but they don't know what. The blank page isn't just a writer's problem.

That's why ABI stopped being just a place to ask questions. Today, it organizes guided learning paths, proposes exercises, runs brief quizzes to check comprehension, and allows students to pick up right where they left off—even if three weeks have passed. Asking questions itself turned out to be just as important as answering them.

The Mentor Is Still Human

On the instructor's side, something less visible happens. Those same interactions reveal to the teacher which concepts confuse students most and where learning paths tend to stall.

And human intervention is part of the core design, not a patch. If a student requests to speak with a professor, the query is escalated directly. If a response falls short of defined standards or touches on a topic the system shouldn't handle alone—such as grades, passing criteria, or a sensitive concern—that specific answer is held: a teacher reviews, corrects, or rewrites it before the student receives it.

This isn't about reducing learning to a metric or leaving sensitive decisions to a machine. It's about arriving on time: knowing who is falling behind while there's still time to act.

Who Benefits Most

First, those with the least margin. Not all students have equal opportunities to ask questions: the student working night shifts, the one commuting two hours, the shy one, or the student who realizes they're stuck on Sunday afternoon when sitting down to study. A teacher can't always be available. An assistant can. It’s not the same, but it’s infinitely better than nothing.

Second, those who learn at a different pace. In a classroom, one explanation serves thirty people: someone needs to hear the concept again, more slowly, while someone else already got it and wants to push further. With ABI, a student can ask for the same idea with more detail, with another example, or connected to a previous topic, without having to choose between holding up the whole class or falling behind in silence.

And third, something I hadn't anticipated: the course itself.

When Students Reshape the Syllabus

While reviewing queries, a pattern emerged: many students were asking about agent-to-agent communication—how one AI system coordinates with another—a topic barely mentioned in the original course materials.

I didn't spot this through teacher's intuition or by reading about tech trends. I spotted it because the questions kept coming. Today, that topic is fully integrated into the syllabus.

That is probably the most interesting takeaway of the whole project. Grouped together, student queries create a map of real struggles and genuine interests. Not what you assume as an instructor, but what actually surfaces in practice. That insight always existed, but it used to get lost in hallways, group chats, or unspoken doubts.

The Obvious Objection

Isn't this just a shortcut for cheating?

It's the first question that comes up, and it deserves an honest answer. ABI isn't designed to solve exercises for students, but to explain them: it relies strictly on course material, cites sources, and directs students back to the text rather than replacing it.

That said, we shouldn't be naive. No tool guarantees proper use on its own. What changes the outcome is how you evaluate. If your sole assessment is a take-home assignment, the problem existed long before generative AI. If class time is used to discuss and test what was studied, an assistant that better prepares students for that session works in your favor.

What Technology Can't Replace

There is something ABI cannot do, and it should be stated plainly: it can't read a confused expression in the third row. It can't sense when one question is hiding another. It can't challenge a student who is slacking off or facilitate the collective conversation where one person's question clarifies things for everyone else.

That takes a teacher. And it always will. The value of an agent like ABI isn't in replacing that role, but in freeing it up: if repetitive questions find answers outside the classroom, class time remains open for what can only happen inside it.

The Bottom Line

There’s a lot of talk about integrating AI into education. There’s far less talk about why. Technology doesn't improve learning simply by being present; it improves learning when it solves a specific, named problem: a student unable to ask a question, a weak grasp of a concept, or a topic missing from the syllabus.

The scene from the beginning hasn't disappeared. Class still ends, and two or three students still walk up to the desk—always the same ones. The difference is what happens to the other twenty-eight.