Nearly every lesson reviewed, up from about 20 a month
In four months, from late January to late May 2026, the school went from about 20 observed lessons a month to a review of nearly every lesson it taught. Every teacher got Lansy AI as an assistant, and the leadership saw the whole school for the first time.
4,796
lessons with a full review
25
criteria in every review
51
teachers at the school
Every connected lesson was recorded and reviewed by Lansy AI against the school’s own criteria.
A private school, primary to secondary school
- Format
- Extended-day school, 8 a.m. to 7 p.m.
- Classes
- 22
- Teachers
- 51
- Promise to parents
- Individual attention and consistent quality in every lesson
The school promises parents individual attention and consistent quality of teaching. With several dozen teachers, the principal cannot know what happens in most lessons, so management rests on a sample rather than on data.
The school runs an extended day, primary to secondary school. Traditional methods cannot check every lesson, even with daily learning walks.
This year the school decided to change how it manages quality and to leave its teaching methods as they were.
Coordinator visits covered a few lessons a month, and Lansy AI reviewed nearly all of them
Before Lansy AI, the routine was familiar. The academic coordinator sat in on a lesson for 40 minutes, wrote a report, and prepared feedback. It is good practice, but its reach is limited by one person’s working hours.
About 20 lessons a month
Scheduled observations by the academic coordinator and the principal.
4,796 lessons reviewed
Nearly every lesson the school taught during the pilot.
The school stopped judging itself by a few showcase lessons and, for the first time, saw nearly the whole picture of its teaching.
A scheduled visit turned the lesson into a showcase, and feedback came days later
The problem lay in the observation model rather than in the coordinators. When teachers know the coordinator is coming, the lesson becomes a showcase. Feedback arrives a few days later, when the details are already forgotten. Assessment ends up living apart from the teacher’s growth.
The school wanted a different cycle, and that became the task for Lansy AI.
The cycle the school wanted
The lesson
Every connected lesson is recorded.
A review within minutes
Lansy AI scores the lesson against the school’s criteria.
Specific recommendations
What to change, with the moment of the lesson it comes from.
The next lesson
The teacher tries the change as soon as the next day.
A new check
The next review shows whether the change worked.
Teachers accepted the reviews faster than the school expected
The school did not launch across the board. Seven volunteer teachers went first. The rest saw real reviews and only then joined.
The most accurate words came from a language teacher rather than from the administration: “I kept putting off opening my first review because I expected to find a list of mistakes. I opened it, and it was simply a conversation about my lesson. Now I read every one.” That became the main principle of the rollout.
Every conclusion comes with quotes from the lesson. Instead of a verdict like “the explanation was weak,” the teacher sees the exact exchange the conclusion is drawn from, along with what to change as soon as tomorrow. When an assessment can be checked, resistance disappears.
What teachers said about the reviews
Chemistry teacher
“We looked at the lesson reports. The analysis is excellent, the wording is as tactful as it gets, the recommendations are great: no filler, everything to the point, clear and well structured.”
Maths teacher
“I agree with the whole review, both the strong points and where there is room to grow. That’s what you want from feedback.”
Physics and technology teacher
“I now start preparing for a lesson by rereading the review of the previous one. It’s a new habit, and a good one.”
A review shows what a lesson rests on and what to try next time
Every lesson is scored against the school’s 25 criteria and gets one score out of 100. Beyond the score, the review shows what the lesson rests on, what weakens it, and what to try in the next lesson. Here is one lesson of the pilot in full.
A medieval state through a board game
Teacher · Social studies · Year 6
Overall score
91 /100
Key terms
Used in play
Timing
Stages on schedule
Reflection
At the end
Involvement
Part of the class watching
Summary
The terms work instead of being defined. The key concepts of the topic, such as power, domain, and council, are heard in the pupils’ own moves: to make a move, a pupil has to apply the term. The score rests on practice and timing. The rules were learned through action, the stages fit the schedule, and the lesson ended with reflection. The score is lowered by arguments about the rules, by part of the class staying spectators, and by a conflict at the peak of excitement.
What held the score and what lowered it
Part of the class stays spectators, and arguments about time and the rules flare up at the peak of the game.
Better approach
Introduce the roles of chroniclers and judges: they bring the spectators into the game. An hourglass for each move settles the disputes about time. End with what each player liked in the opponent’s play: which of the opponent’s moves was the strongest.
Game-based activities fall to 18.6% by upper secondary school
The most interesting findings concerned patterns across the whole school rather than individual teachers. The first is the share of game-based and interactive activities, which falls with every age group.
Game-based and interactive activities, share of available points
51.9%
The highest share of the three age groups.
30.0%
Games and interactive work give way after primary school.
18.6%
The lowest share of the three age groups.
Three qualities do not depend on age: the teacher’s subject expertise at 96.5%, clarity of explanation at 94.9%, and logic of presentation at 88.4%.
The weakest area turned out to be shared by the whole school
31.9%
The average result on the criterion for how a lesson ends.
44.2%
The school loses the last minutes of almost half its lessons.
The subject matters more than general advice
No recommendation fits every subject. The reviews showed what works best in each group of subjects.
Maths and physics
Need short, step-by-step explanations and work at the board.
Literature
Discussion and a link to the pupil’s own experience have the greatest effect.
Foreign languages
Games work better for consolidating new vocabulary.
Sciences
Observation, experiments, and guiding questions engage pupils more.
A single visit found nothing wrong, but dozens of reviews showed the explanations ran too long
Parents of a Year 8 class complained that the maths teacher “can’t explain anything.” The academic coordinator visited a lesson and saw nothing to remark on.
Lansy AI found something else: the explanation was good but too long. The class’s attention dropped, and the second half of the monologue turned into noise for the pupils.
The teacher needed shorter explanations. That pattern shows only across dozens of lessons by the same teacher. Lansy AI also keeps a profile of every class.
Year 8 class profile
Productive but noisy, needs structure · 27 January – 24 February 2026
The class is highly social and needs to talk, which often shows as constant background noise. The pupils are not aggressive but are easily distracted, and their discipline and engagement depend directly on the teaching style and the lesson format. The class shows high potential with interactive, hands-on, and dynamic methods but loses focus during monotonous work. The informal relationships within the class call for confident management and clear, structured instructions from the teacher.
Average score
74%
Lessons reviewed
22
Skills
Consistently strong
Interactive, hands-on, and dynamic lessons
Growing
Lansy AI took over observing lessons and left the coordinator the work that needs experience
The administration began to make decisions based on recurring patterns. Instead of general recommendations, it holds targeted department meetings and sees signals of an emotional dip or lost attention in a class.
Lansy AI did not replace the academic coordinator. It took the endless observing of lessons off the coordinator’s plate and left the work that requires experience: discussion, support, and the teacher’s growth.
The case rests on every lesson of the pilot
The case is based on an analysis of all the lessons the school taught during the pilot. Lansy AI scored each one against the same criteria, so the school could compare age groups and subject areas, 18 subjects in all.
A separate study of the same lessons looks at which methods keep pupils engaged depending on the year and the subject: which words are heard in the most engaging lessons, whether the teacher’s sense of a lesson matches the data, what can make a class’s engagement collapse, and how the same class engages in different subjects.
The school will judge teaching quality after a full academic year
A full year before conclusions
Four months is too early for final conclusions about the growth of teaching quality. The school will wait for at least a full academic year.
Decisions on data
One conclusion already holds. With almost 5,000 lessons in view instead of a random handful, the principal runs the school on data.
The last minutes of every lesson
A summary at the end of the lesson becomes the focus of next year’s professional development.
See Lansy AI on your school’s lessons in three days
Start with one class or 5–10 lesson recordings. Your team sees the first reviews, checks their accuracy on its own lessons, and decides what to change in the next ones.