TutorTracer uses large language models to read lesson transcripts the way an experienced reviewer would. Every flag and rating points back to the words that caused it, and the final decision always rests with your team.
Free for 14 days or 400 lesson hours, whichever comes first. No card needed.
Every issue carries a verbatim quote and a timestamp. Every quality rating cites what happened in the lesson. If you can't see why, it isn't useful.
If a transcript is too short, garbled or one-sided to judge, the quality assessment says “insufficient evidence” instead of guessing, and it is excluded from averages.
TutorTracer is decision-support. It doesn't discipline tutors, contact families or close issues. Your team does.
Plain-English Issue Guardrails, a custom quality rubric and your own curriculum and policy documents shape what it looks for.
Your lessons are processed through AI providers' APIs under terms that stop the provider training models on them. We don't train foundation models on customer data either.
Transcript excerpts are sent to AI providers for analysis, transiently, under terms that stop them training on your data.
The transcript and lesson details arrive from your platform once the platform has produced them.
Tutor and student are identified from the platform's roles, or from the conversation, cross-checked against who speaks most.
The transcript is checked for the four issue categories. Each issue gets a quote, timestamp, severity, context and explanation.
The lesson is rated on each rubric dimension, with evidence, or abstains.
The lesson is indexed so the assistant can search it.
Everything depends on knowing who said what. Where the platform labels exactly one teacher, TutorTracer uses it. Otherwise it works it out from the conversation and checks the answer against who talks most.
The TutorTracer assistant answers questions about your lessons, tutors, students and flagged issues, using the lessons it has indexed, and links to the lessons behind each answer.
AI outputs can be incomplete, wrong or mis-attributed. They can miss serious issues and flag innocent ones. Treat them as leads to check.
If your platform doesn't produce a transcript, for example Lessonspace with transcription switched off, the lesson can't be analysed.
No automated action is taken against tutors or students. Outputs are not findings of fact.
We use large language models from third-party AI providers, accessed through their APIs. We choose and test models for each step of the analysis and change them as better ones become available.
No. We use API terms under which the provider is contractually restrained from training its models on your content, and we don't use customer data to train foundation models.
Accuracy depends on transcript quality, the subject and your settings. Every output shows its evidence so you can check it quickly, and guardrails let you cut out flags you consider noise. The best way to judge is on your own lessons, which is what the free trial is for.
Yes: plain-English Issue Guardrails, your own quality rubric, and checks against your own curriculum and policy documents.
Connect Lessonspace in a couple of minutes and TutorTracer analyses your last seven days of lessons straight away. Or book a walkthrough and we'll show you the product on demo data first.
Free for 14 days or 400 lesson hours, whichever comes first. No card needed.
