Cohort answers them in the context of the actual lesson, in the learner's own words, and then tells your instructors which two students genuinely need a human this week.
Works with Teachable, Thinkific, Moodle, Canvas and Circle
You're not subtracting it again, you're carrying it forward. In the week 3 sheet the opening balance is week 2's closing balance, so it appears once as a starting point, not as an outflow.
Look at cell B14 in the workbook from the lesson: it references B32 from the previous tab. If you change week 2's closing figure, week 3's opening moves with it.
Cohort is grounded entirely in the material you uploaded. If the answer is not in your course, it says so and passes the question to a human.
Video transcripts, slides, workbooks and your existing FAQ, indexed down to the timestamp.
Responds in the thread the learner asked in, citing the lesson and minute it came from.
Tracks who has stalled, who is re-watching the same eight minutes and who has gone quiet.
Hands the instructor a short list of learners with the context already written up.
Not a churn score. A named list with the reason attached, so the intervention is specific rather than a nudge email.
Three limits, set in the engine rather than in a prompt, because a teaching assistant that oversteps costs you more than it saves.
If a question falls outside the uploaded material, Cohort says it doesn't cover that and routes it to an instructor. It does not fill the gap.
It will explain a concept four different ways. It will not produce the answer to a graded exercise, however the question is phrased.
Learners are told they're talking to an assistant and can ask for a human at any point. The handover is one click and it always works.
Professional certification courses, 200 to 1,400 learners per intake.
We index the material, run it alongside your instructor for a fortnight, and you compare its answers against theirs before it speaks to anybody unsupervised.