Skip to content

Product overview

The problem it solves

A small IELTS-focused English center runs on spreadsheets and messaging apps: class rosters in one sheet, attendance in another, tuition in a notebook, and practice tests handed out on paper or PDF. Grading Writing and Speaking is slow and inconsistent. KATE General replaces that with one system that handles both running the center and delivering AI-assisted practice, so staff spend less time on admin and students get instant, structured feedback.

Two products, one platform

graph LR
  subgraph "Center Operations (the 'OS')"
    A[Classes] --- B[Students]
    B --- C[Teachers]
    C --- D[Schedule & Attendance]
    D --- E[Tuition & Cashflow]
    E --- F[Reports]
  end
  subgraph "AI IELTS Practice"
    G[Question bank] --- H[Assignments/Homework]
    H --- I[Reading · Listening · Writing · Speaking]
    I --- J[AI grading + teacher approval]
    J --- K[AI Tutor]
  end
  F -. shared roster, one login .- G

The two halves share the same accounts, roles, and login. A student's practice results feed their profile in the operations side; a teacher's grading workflow lives in both.

Who it is for

Audience What they do here
Manager / Ban điều hành Runs the center: classes, enrollment, schedule, attendance, tuition, cashflow, reports, and academic setup (question bank, assignments, materials, AI settings).
Teacher / Giáo viên Teaches and grades: own classes and students, grading queues (Writing/Speaking), question authoring, homework assignment.
Student / Học viên Practices assigned homework only: does tests, gets instant results and AI feedback, tracks band progress, uses the AI Tutor when enabled.
Parent / Phụ huynh Planned (Phase 2) — a read-only portal for schedule, scores, and fees.

Design principles

Built for non-technical operators

The center is run by teachers and administrators, not software engineers. The product favors one clear task per screen, Vietnamese UI copy for all operations features, a calm neutral palette with a single purple accent, and no exposed technical jargon (users never see model names like "Whisper" or "Gemma" — only "AI").

  • Two-mode navigation. A navbar mode for browsing and a focus mode for doing a single task (taking a test, grading an essay, marking attendance) without distraction.
  • Homework-only for students. This is a center tool, not a self-study app. Students take only what a teacher assigns — enforced on the server, not just hidden in the UI. History of completed work stays readable.
  • AI assists, humans decide. Every subjective AI band is provisional until a teacher approves it.
  • Course-based, not subscription. Revenue is the tuition ledger per course plus cashflow — there is deliberately no online payment gateway and no per-student subscription plan.

Scale & economics

  • Target size: ~150 active students, one center.
  • Operating envelope (from the product proposal): a flat monthly budget covering roughly 1,000 AI Writing gradings, 6,000 Speaking minutes, 6,000 AI Tutor turns, and 50 GB storage per month. These are enforced as hard caps in the platform so a single user cannot exhaust the shared budget — see AI features → Quotas.
  • Infrastructure cost: Cloudflare Workers/D1/R2/KV at this scale sit within low-cost tiers; there is no server or database host to pay for separately. AI usage is the main variable cost, which is why it is metered and capped.

Roadmap at a glance

  • Phase 1 (now): the eight management modules, two practice modes (self-check with AI feedback + teacher-assigned homework), AI Writing/Speaking grading with approval, AI Tutor, role dashboards.
  • Phase 2: parent portal, mobile app, push notifications, advanced reports, Zalo/SMS.
  • Phase 3: handwritten-Writing OCR, deep pronunciation analysis, white-label, multi-branch.

See Status & gaps for an honest current-state assessment against this roadmap.