The AI is in the authoring. Not in the classroom.
A LessonSlate lesson reaches a learner as a video file and a scored quiz. There is nothing to chat with, nothing that answers back, and no student data collected — by us or by the lesson. The AI did its work upstream, before a teacher ever reviewed and assigned it.
Schools where the teaching already happens on video
In a virtual school, an online course, or a credit-recovery programme, instruction is already delivered as a video plus an assessment. That is the normal shape of the work there, not a compromise. Nobody is being replaced — the teacher is a person managing a caseload, deciding what each learner gets next.
The bottleneck in those programmes is not staffing. It is that video cannot be made fast enough to match what a cohort is actually failing on, so a fixed library gets assigned whether or not it fits the misunderstanding in front of you.
LessonSlate changes what is affordable to make. It does not change who decides.
The teacher reads the assessment data, picks the problem type, presses the button, watches the result, and assigns it. Automation sits only in the middle — where the animation and the voice work used to cost a studio.
That means a lesson can be made in response to a quiz result: the exact problem type a group got wrong this week, delivered as a new module next week, in the same voice and format as everything else in the course. Human judgement at both ends. A machine doing the drawing.
Virtual & online schools
The lesson library keeps pace with what this term's cohort is getting wrong.
Sub & absence libraries
A district-wide bank of worked examples that a substitute can assign without teaching the topic.
Two different things get called "AI in schools"
Districts are increasingly restricting AI that students interact with directly — chatbots, tutors, companion apps and monitoring tools. Those policies are written about software a child talks to. LessonSlate is on the other side of that line, and it is worth being precise about which side.
What the AI does not do
- It does not talk to the student. There is no prompt box.
- It does not answer questions, adapt, or remember anyone.
- It does not read, scan, or monitor student activity.
- It does not run at all while the lesson is being watched.
- It never receives a student's name, image, voice or work.
A student watching a LessonSlate lesson is doing exactly what a student watching any instructional video does. The entire difference is upstream, in how the video got made — which is the part a teacher reviews before assigning it.
Four worries about AI in classrooms, and where we actually sit
These are the objections that drive current policy. We would rather answer them in writing than have you find out on a call.
| The concern | What it is about | LessonSlate |
|---|---|---|
| Cognitive off-loading "students stop thinking" |
A learner handing the thinking to a chatbot, and losing the ability to do it themselves. | There is nothing to hand it to. The lesson shows the thinking: the on-screen student hesitates, guesses wrong, and is corrected. Watching somebody reason through a worked example is one of the oldest methods in maths and science teaching. We did not invent a pedagogy — we made an old one cheap enough to produce one per problem type. |
| Companion chatbots software that acts like a friend |
Systems that simulate a relationship with a child during years when that matters most. | The on-screen student is a character in a video, like any character in any educational film. It never responds to the viewer, never remembers them, and cannot be spoken to. |
| Surveillance monitoring what students write and do |
Scanning tools that read student messages, files and browsing, on and off campus. | We are not a monitoring product and have no monitoring feature. We receive no student data of any kind, so there is nothing here to scan with or to hand over. |
| Deepfakes and likenesses | Tools that generate images or voices of real students. | No learner's image, name, voice or written work is ever an input to anything we run. The personas are invented characters with licensed synthetic voices. |
The six questions you are going to have to answer about us
If you are the person who has to fill in a technology-review form, these are the answers. There is a printable one-page version at the bottom.
-
Where does the AI run?
At authoring time only, on our machines, before anyone assigns the lesson. No model runs during playback. The finished lesson is a video file, a caption file and a quiz — none of which call anything, anywhere.
-
What student data do you receive?
None. There are no student accounts and no logins. The package we hand you makes no outbound network calls at all — it speaks only to your own learning platform, through the standard interface that platform provides, to write a completion and a score into your own gradebook. The record stays on your side of the wall. Nothing is sent to us, so there is nothing for us to lose, sell, or be compelled to produce.
-
Who checks the content before a student sees it?
A machine, then a person, in that order. Before a lesson can be published, an automatic checker recomputes every step of the working and compares the final answer against the answer the problem was set with. If they disagree, the lesson is refused — the publishing script exits with an error rather than shipping it.
This is not theoretical. A lesson once divided 0.7818 by 204.23 and wrote 0.0038289 instead of 0.0038280. That 0.02% slip carried through to a wrong final answer, and every other stage of the pipeline passed it. The arithmetic check caught it before it went anywhere. After that, a human reviews the lesson and approves it.
-
Where do the quiz's wrong answers come from?
From 661 hand-written notes recording what learners actually get wrong on each concept — each one a real mistake paired with its correction. A question is that pair turned around: the correction becomes the right answer, the mistake becomes the option sitting next to it.
No distractor is invented by a model. That is the difference between a wrong answer that tells you which misunderstanding a learner is carrying and one that is plausible-sounding filler.
-
What does it deliberately not do?
It cannot answer a question the lesson did not anticipate. It does not branch — a wrong quiz answer is not routed to a different explanation; you would assign a different lesson yourself. It does not adapt to an individual learner in real time, and it holds no profile of anyone. It replaces the first explanation, not the tutor.
-
Where does the source material come from?
Every problem carries a provenance record that travels with the lesson. A lesson built from your own course material is stamped as yours: it is delivered to you and is not added to any catalogue we license to anyone else. The publishing check verifies that stamp survived to the finished file, rather than trusting that it did.
Where we would not be a good fit
Better you find this here than after you have built a rollout around us.
A blanket ban on AI-generated material
Some policies object to the origin of the content, not to where the software runs. Correct scoping does not help there, and we will not pretend otherwise. What we can offer is full disclosure and a written human-review record — but if the rule is absolute, we are not your vendor.
Hard screen-time limits in early grades
Some restrictions are grounded in screen-time policy rather than AI policy, and bar screens outright for the youngest learners. LessonSlate is a video product, so that rule catches us. In practice our subjects start at Algebra 1, so this rarely comes up — but it is a real limit, not a technicality.
You need it to answer live questions
It cannot. A lesson anticipates the common question and shows the common mistake; it cannot respond to a new one. If what you need is a tutor a student talks to, that is a different product — and, in most districts right now, a restricted one.
We say it is AI-made, early and plainly
Lessons are authored with AI assistance and we state that on the lesson, in the description, and here. We are not going to soft-pedal it, bury it in a terms page, or let a reviewer discover it later. A vendor who is vague about this in 2026 is telling you something.
What we will argue for is the distinction that actually matters to a school: whether the software is something a child uses, or something a teacher used to make a lesson. Those are different products with different risks, and they deserve different rules.
- AI-assisted authoring, disclosed on the lesson itself
- No student-facing model, at any point, in any subject
- No accounts, no logins, no cookies, no trackers
- No outbound calls from the package we hand you
- A person approves every lesson before it ships
- Captions on a real subtitle track, keyboard-navigable quiz
Send us your policy and we will answer it directly
Tell us which learning platform you run and what your technology review asks for. We will answer against your form rather than hand you a brochure.
Get in touch