Skip to content

The journey of one lesson

Here is the entire path of a lesson: from its start to the finished report in the admin panel. First we show it through a live example, then the same flow as a diagram, and then strictly step by step. No technical details — just what happens and in what order.

The story of one lesson

Names and figures are made up, for illustration.

Monday, 14:00. In classroom No. 3 of the learning center, the IELTS B-12 group (level B2) starts its lesson. It is taught by teacher Dilnoza R. As always, there is a camera in the room — it records the lesson (picture and sound). Neither Dilnoza nor the students do anything special: the lesson runs as usual.

15:30. The lesson ends, the students leave. At this point human involvement in the "magic" ends — from here on everything is done by ControlAI.

15:31. From the schedule, the system knows: the lesson of group B-12 in classroom No. 3 has just ended. It pulls the recording of this lesson from the center's NVR/DVR on its own — no one uploads files manually.

Then ControlAI analyzes the lesson from three angles at once:

  • listens — performs speech-to-text and works out who spoke (Dilnoza or the students), how much, how many questions and pauses there were, what share of the lesson was in the target language;
  • watches — from the video, counts how many students were in the room and how many arrived late (this part never delays the report and may come later);
  • checks against the methodology — fills in the center's rubric: did Dilnoza deliver the planned topic the way it is supposed to be done.

16:30, about an hour after the lesson. The report is ready and appears in the admin panel. Everyone sees their own view:

  • The owner opens the dashboard and sees the B-12 lesson card: the teacher spoke 68% of the time, the students — 32%, 24 questions, English 88%. The video metrics — 14 out of 16 attended, two arrived late — will follow a bit later, by the evening digest. Next to it — the "lesson DNA" (a colored ribbon showing how the time was distributed) and a couple of flagged moments.
  • The methodist (academic manager) opens the filled-in rubric, quickly checks the AI's scores and, if needed, corrects them.
  • Dilnoza logs into her personal account and sees her own statistics and a short summary of the lesson — not a "verdict from above," but data that helps her understand what to improve.

No one re-watched the recording. No one filled in a form by hand for hours. The lesson simply turned itself into clear numbers and a report — and so it goes for every lesson of every class.

The same path — as a diagram

flowchart TD
    A["Lesson is running — camera records video and sound"] --> B["Lesson ended (per the schedule)"]
    B --> C["ControlAI pulls the recording from the center's NVR/DVR on its own"]
    C --> D["Lesson analysis"]
    D --> D1["Listens: who spoke, questions, language, pauses"]
    D --> D2["Watches: attendance, late arrivals"]
    D --> D3["Checks against the methodology: the center's rubric"]
    D1 --> E["Finished report — in about an hour"]
    D3 --> E
    D2 -.-> E2["Video metrics — later, they complement the report"]
    E --> F["Admin panel"]
    E2 -.-> F
    F --> G1["Owner"]
    F --> G2["Methodist"]
    F --> G3["Teacher"]

The same path — step by step

  1. The lesson is running. The camera in the room records video and sound — as usual, without any human involvement.
  2. The lesson has ended. From the schedule, ControlAI knows when and where the lesson finished.
  3. The recording is pulled. The system downloads the lesson recording from the center's NVR/DVR on its own — nothing needs to be uploaded manually.
  4. Sound — speech. ControlAI "listens" to the lesson: it transcribes what was said and separates where the teacher spoke and where the students did.
  5. Video — the room. Separately, without delaying the report, it "watches" the picture: how many students are in the room, how many arrived late (these metrics may come later than the speech ones).
  6. Computes the metrics. Talk-time balance, the number of questions, pauses, the target-language ratio, punctuality, attendance — all in numbers.
  7. Checks against the methodology. Fills in the center's rubric and checks whether the planned topic was covered.
  8. Writes the report. Turns the ready numbers into a short, clear text.
  9. Shows it in the admin panel. The result appears in the panel — everyone, under their own account, sees what concerns them.

What if there was no lesson?

There are two cases here.

The lesson was canceled in advance. If it is already known that there will be no lesson (a holiday, a substitution, a group cancellation), this is marked in the schedule in the admin panel in advance — and ControlAI simply does not take this lesson into processing.

The lesson simply did not happen. If the lesson remained on the schedule but no one showed up or the room is empty, ControlAI notices this on its own: there is no speech in the recording and no one in the room — it does not waste time on a full analysis, marks the lesson as no-show, and does not charge for it. The owner sees the flag and understands what happened.

How long to wait for the result

The main lesson report (speech, questions, methodology) is ready in about an hour after the lesson ends. The video metrics (attendance, late arrivals) may be added later — within a day, usually by the evening digest; they complement the report and never delay it.

In centers with a weak internet connection, the recording may be pulled not right away but overnight — in that case the report will be ready the next day. This is agreed during onboarding.

The key takeaway from this chapter: after the lesson, ControlAI pulls the recording on its own, listens to and watches the lesson, computes the metrics, checks against the methodology, and in about an hour shows a finished report in the admin panel — separately for the owner, the methodist, and the teacher. People do not re-watch recordings or fill in forms by hand.

→ Next: 04. What ControlAI can do — a full list of what the system sees and measures.