AI assistant for an online school
It answers students from the course materials, restores access and reminds about deadlines, without doing the assignment for them. The curator gets only what needs a person.
- Buildfrom $1,700
- Timeline2–4 weeks
- Supportfrom $570/mo
Who this suits: Schools where one curator runs several cohorts and every cohort brings the same questions about access, recordings and deadlines.
What is in the first version
- course answers
- access recovery
- deadline reminders
- curator handover
The figures above are the package this sits on: AI agent or process automation. Anything past its edges the calculator adds before the work starts, not after.
How the path goes
- 01
A question from a known student
The assistant sees the cohort, the open module and the payment, so it answers with a fact.
- 02
Answers lead back into the course
It answers from the materials and links the place in the lesson, and gives no finished solution.
- 03
The deadline announces itself in advance
The reminder goes only to those who have not handed in; whoever still misses reaches the curator with context.
What decides whether this works or annoys people
The assistant that does the homework
A hint towards the step and a link to the lesson yes, a finished solution and test answers never.
Access, payment and a fault differ
The assistant names the real cause and never opens access itself out of politeness.
One outage arrives as many questions
A repeating complaint is a trigger: the assistant stops explaining, alerts a person and tells everyone the same thing.
What to measure once it is live
- share of questions closed without a curator
- time to an answer at night
- access messages per cohort
- lessons that generate most questions
Why this comes out faster
The shape of this one is known: the states, the edge cases and the things that usually go wrong have been decided before. Nothing here is a template, and the saving is not in your half of the work. It goes into your process, your content and the systems this has to talk to, which is the part nobody can have solved in advance.
A likely stack for this
Picked against the task when we scope it, not decided in advance. This is the shelf it usually comes off.
- Assistant and knowledgePythonFastAPIOpenAIAnthropicpgvector
- Access and dataPostgreSQLRedisS3Docker
- Channels and schedulingTelegram Bot APIn8nTemporalGrafana
Price it yourself, right here
Five steps, and you can see the number without leaving a contact. The estimate accounts for the kind of work, what you already have and what it will need inside.
These bills do not come from us
- AI model calls at the provider’s rate
- The course platform subscription
- Video storage and delivery
- The email service for reminders
Asked before the first call
- Will it do the assignments for students?
- No. It gives a direction and a link to the lesson, never answers test questions, and every attempt to get a finished answer stays in the log.
- Where do its answers about the course come from?
- From your own materials, versioned by cohort, so each student gets the version they are studying. What is not in that source it hands to a curator.
- Can it open access by itself?
- A lost link and a sign-in yes, that is reversible. Extending a deadline or opening a paid module goes through a person’s confirmation.