How it works
When a learner asks a question, the platform finds the most relevant passages from the course and sends them to a language model with the question. The model writes an answer from those passages. Without that step the tutor is a general chatbot that happens to sit next to your course.
Each exchange uses model capacity, measured in tokens. A longer course excerpt or a longer conversation costs more per question.
What to check
These seven checks separate a tutor you can put in front of paying learners from a demo.
Check | Why it matters |
|---|---|
Grounding | Answers should come from your course content, not from the open web. |
Moderation | Learner questions and uploaded documents should be screened before they reach the model. |
Usage limits | Each learner and each workspace needs a cap, or one person can spend the whole budget. |
Conversation memory | The tutor should remember the session so a follow-up question makes sense. |
Staff visibility | Instructors should be able to see what learners ask, to find gaps in the course. |
Availability | Ask what learners see when the AI service is down or the cap is reached. |
Data handling | Ask which model provider processes learner questions and whether they are used for training. |
Where a tutor helps, and where it does not
A tutor is good at the question a learner would not ask in front of a group: a definition, a worked example, a second explanation of a step. It answers at night and between live sessions.
It is weak at judgement. Feedback on a project, a career question or a disputed grade still needs a person, and a good platform makes it easy to hand over.
How Perfectum AI does it
In Perfectum AI the tutor sits inside the course player and answers from the course’s material. Questions are moderated, each learner has a usage allowance, and each workspace has its own budget. When a limit is reached the learner is told so plainly.
Atin Singh is the founder and CEO of Perfectum AI. About the company