Almost every school software vendor selling in Karnataka claims Kannada support. Very few will tell you what fraction of the product is actually in Kannada, or what happens on a screen that has not been translated. This page answers that with specifics, because a school buying on a vague claim discovers the truth in week two.
What “Kannada support” actually means in software
The phrase covers three separate things, and vendors let them blur together deliberately:
- The staff interface — menus, buttons, table headers, form labels and error messages. This is the expensive one: translated once, then maintained forever, because every new feature adds strings.
- Messages to parents — WhatsApp replies, SMS, push notifications. Cheaper, and the layer most vendors mean when they list ten languages.
- Documents — report cards, receipts, transfer certificates, circulars. Different again, because it depends on font rendering in a PDF, not on a translation file.
A vendor answering “yes, we support Kannada” without distinguishing these has told you nothing.
How much of the interface is really in Kannada
EdunodeX ships translation bundles for English, Hindi, Kannada and Tamil. Measured against the English source, the Kannada bundle currently covers about 1,291 of 1,545 interface strings — roughly 84%.
The covered areas are the ones a school office lives in daily: students and guardians, admissions, fees, attendance marking, academics and teaching plans, homework, transport, reports, staff, settings, navigation and the dashboard.
What happens to the remaining 16%: an untranslated string falls back to English on that screen. It does not render blank, and it does not show a raw key name. In practice a Kannada user sees a mostly-Kannada screen with occasional English labels, usually in newer or more specialised modules.
That is a real and useful level of coverage, and it is also not 100%. Both halves matter. If you are comparing vendors, ask each one for the same number — translated strings over total strings — and watch how quickly the conversation changes.
A Karnataka school gets Kannada by default
There is a difference between a language being available and a language being what your staff actually sees. Most non-technical school staff never open a language toggle.
EdunodeX resolves this by state: a tenant whose state is set to Karnataka defaults to the Kannada interface for any user who has not explicitly chosen a language. Tamil Nadu defaults to Tamil, the Hindi-belt states to Hindi. States with no real translation bundle deliberately stay on English rather than switching to a language code that would silently fall back — a default that pretends to be regional and then shows English everywhere is worse than an honest English default.
An individual user can still switch at any time, and their choice takes precedence over the state default.
What the parent actually receives
The parent-facing layer works differently from the interface, and it reaches further.
A parent who writes to the school’s WhatsApp number in Kannada is answered in Kannada. So is a parent writing in Hindi, Tamil, Telugu, Malayalam, Marathi or Hinglish — the intent classifier is multilingual, so a parent can both ask and be answered in their own language.
Two design decisions in that flow are worth knowing about:
- Facts are computed in code and preserved verbatim. Fee amounts in rupees, due dates, marks, attendance percentages and names are produced by the system and passed through unchanged. The language model rewords the sentence around them; it never generates the number. A fee figure cannot be hallucinated or rounded in translation.
- The reply speaks in the school’s voice. Parents are not told they are talking to a bot, an assistant or an AI.
If the model is unavailable — or the school has no AI credits left — the reply falls back to English rather than being dropped. A parent always gets an answer.
Consent, opt-out, and one honest gap
Outbound WhatsApp is where Indian schools most often get themselves into trouble, so the controls matter more than the language.
Business-initiated messages use Meta-approved templates only. Freeform outbound text is refused at the backend service layer before the payload ever reaches Meta, which is what keeps a school’s number out of policy trouble.
Opt-in is tracked per guardian, with the source recorded — admission form, self-serve, manual entry or import — and opt-in and opt-out timestamps. A send to a guardian who has not opted in is rejected with a consent error before any call to Meta is made.
Parents can stop messages by replying STOP, and START resumes them. Here is the honest gap: the keyword handlers currently recognise English plus Hindi aliases — बंद, रोको, रुको for stop and पुनरारंभ for start — but not Kannada equivalents. A Kannada-speaking parent who types the Kannada word for “stop” will not be automatically opted out; the school has to do it from the admin side. That is a real limitation and worth knowing before you tell parents to reply STOP in a circular.
SMS is a separate channel with separate rules. It is off by default per school and fails closed until an admin enables it, and every send uses a pre-registered DLT template — free-form SMS is refused. Your SMS goes out in whatever language you registered with the template.
When a parent’s number is not on file
A recurring problem is that the number on record is the one the father gave at admission four years ago, while the mother is the one who actually messages the school.
When an unrecognised number contacts the school’s WhatsApp, EdunodeX runs a self-service verification wizard rather than a flat refusal: it asks for the student’s admission number, then the date of birth, then a photo of the parent’s ID proof — a school ID card or a recent fee receipt. Aadhaar is deliberately not accepted, because collecting it here would create a data-protection liability the other documents do not. Failed matches at either step return the same generic message, so the flow cannot be used to fish for whether an admission number exists.
That produces a link request in a staff queue, where an admin approves it against a specific contact role — father, mother or guardian — with a conflict warning if that role is already linked to a different number. Uploaded proof files are purged automatically after review.
The parent app, and the teacher who is also a parent
Alongside the browser-based parent portal there is a native Android parent app. Onboarding is by school code — an app link, a QR code or a magic link — with phone or email OTP login and an optional biometric app lock. Inside, a parent gets a per-child hub covering attendance, marks, fees with hosted checkout, payment history, leave requests and report-card PDFs, plus teacher chat and an AI tutor.
Two behaviours matter specifically in Karnataka’s larger schools:
The dual-role case. A teacher whose own child studies at the same school messages from one number. The system resolves both identities up front and routes by what was asked rather than first-match-wins: a fee or leave question goes to the parent handler; a marks-entry question goes to the staff handler. Anything ambiguous prompts a one-tap “Staff or Parent?” choice, remembered for the session. In parent mode the reply is limited to the sender’s own children and never includes school-wide aggregates.
Quiet hours. Push, SMS and WhatsApp broadcasts are held between 21:00 and 07:00 IST by default. In-app notices and email are exempt; emergency messages override the window.
What is not in Kannada
Stated plainly, so you can plan around it:
- Roughly 16% of interface strings are not yet translated and render in English.
- STOP and START keyword aliases exist in English and Hindi, not Kannada.
- Only Hindi, Tamil and Kannada have real interface bundles. Other language codes fall back to English rather than being partially translated.
- Report cards, receipts and certificates follow the template your school configures. Ask for a sample rendered in Kannada before you assume it.
None of that stops a Karnataka school running its office and its parent communication in Kannada today. It does mean you should buy on measured numbers rather than a claim on a slide — and ask every vendor on your shortlist for the same figures.