Hand over everything a teaching assistant should be doing.
Hong Kong students and parents already live in WhatsApp: zero app install, zero learning curve, zero adoption friction. A tutor manages more students without adding headcount, and still knows exactly where each one stands.
Not “a chatbot” — a student-capacity system.
From HK$18,000 for the independent-tutor edition up to institution-wide deployment; renewable annual licence or per-student billing.
One teaching assistant’s job, listed line by line.
The top four are the most time-consuming, most postponed, least pedagogically valuable parts of a tutor’s week — all handed to the system. The bottom two are not handed over, and could not be.
That is the line: the AI serves teachers and owners, not teaching itself. We do not build “AI tutoring”, and no part of this lets a student skip class.
One conversation, two consoles, one report
One system, showing students, tutors and parents only what each should see. Permissions and visibility are set by the institution.
Receive daily practice, submit work, get immediate feedback and the next task.
The system finishes what can be formalised and queues the ambiguous cases back to the tutor.
Class daily and weekly reports give renewal conversations something factual to stand on.
Why we don’t build an app.
Rolling out a new app means clearing two gates: getting students to download it, then getting parents to use it. On WhatsApp neither gate exists.
Who it fits
As an AI teaching assistant: tutors stop spending evenings marking and chasing, and follow-up quality rises at the same class count.
Online, home visits, library or café sessions alike. The light edition builds a professional front and lets you take more students without losing follow-up.
The official account runs under your name — students and parents only ever see your brand, never MNT.
Progress report card
Delivered into the same chat the parent already uses. Format and metrics are defined by your institution — this block is deliberately left blank until the report-card design is finished.
Still to confirm: which metrics appear, how granular the disclosure to parents is, and whether class comparison is shown.
A 6–8 week pilot
Start with one class, one subject. The pilot ends with a results report — if it isn’t a fit, it stops there.
Pick the class and subject; map the tutor’s existing workflow, dispatch rhythm and marking habits.
Open the official account under your name, connect the knowledge asset, dry-run the tutor console.
Students practise on a fixed weekly cadence, tutors handle only undetermined cases, parameters tuned fortnightly.
Completion rates, how weaknesses converged, an estimate of tutor time saved, and whether moving up a tier is warranted.
Upstream and downstream
The assistant marks accurately because a knowledge asset anchored to official schemes sits behind it. That comes first.
LAYER 3 · DOWNSTREAMHow feedback turns into practice actually done — tasks cut short enough to fit an MTR ride.
Bring us one real class.
How many students, which subject, where the tutor’s time actually goes. We’ll return a pilot plan for that class with expected metrics, and say which tier to enter at.