How We Brought Ixorah to WhatsApp — Multilingual AI for Indian Businesses
Auxzon Team
Founding Team
Not a chatbot.
A business system.
"Most AI assistants are built for English-speaking, keyboard-comfortable users. That's not how most Indian businesses actually communicate."
We built Ixorah first for our own website. Then we asked: where do our actual prospects spend time? WhatsApp. Where do Kerala and Indian business users communicate most naturally? WhatsApp — often in Manglish or Malayalam, not formal English. So we extended Ixorah to WhatsApp. This is that story.
Why WhatsApp, not just a website chat widget
WhatsApp has over 500 million users in India. For most SMEs, it is not just a messaging app; it is the primary business communication channel.
- A website chat widget only captures people who are already on your website. WhatsApp meets people where they already are.
- Response on WhatsApp feels personal, immediate, and trustworthy in a way that a website chat widget does not.
The multilingual challenge
Business conversations in Kerala rarely happen in formal English. They happen in Manglish (Malayalam + English mixed), Malayalam script, and English depending on context and who you are talking to.
Most AI systems trained primarily on English data either refuse to respond to Malayalam/Manglish or give incoherent answers. We had to fix this.
What we built: Ixorah detects the language or mix the user is writing in and responds in kind. If someone writes in Manglish, Ixorah responds in Manglish. If they write Malayalam, it responds in Malayalam. If English, English.
The RAG (Retrieval-Augmented Generation) retrieval still happens against the English knowledge base, but the response generation is entirely language-aware.
Lead capture built into the conversation
A traditional chatbot collects a contact form. Ixorah collects intent. Through natural conversation, Ixorah identifies:
- What the user is looking for.
- Whether they have a real business problem.
- Their contact details — all without feeling like filling out a form.
Every qualified conversation gets logged and routed to the team with context, not just a name and number.
WhatsApp Message
User sends a message in Manglish, Malayalam, or English.
Language Detection
System automatically identifies the specific language mix.
RAG Retrieval
Searches Auxzon knowledge base to find accurate answers.
Language-Aware Response
Replies intelligently in the exact language the user used.
Lead Captured
Seamlessly extracts contact info and alerts the team.
What we learned from the first real conversations
- Language switching: Users switch languages mid-conversation — Ixorah handles this gracefully.
- Clarity in comfort: Business queries in Manglish are often more specific and intent-clear than formal English queries. People ask real questions when they feel comfortable.
- Speed is paramount: Response speed matters even more on WhatsApp than on a website. We optimised for sub-2-second first responses.
What this means for businesses
Any business with a WhatsApp presence can now have an AI assistant that speaks their customers' language. Lead capture happens in the channel customers already trust. It's available 24/7, delivers consistent quality, and ensures no conversations are missed.
We built this for Auxzon. We now build it for clients too.
Ready to upgrade your WhatsApp presence?
If your business runs on WhatsApp — and most Indian businesses do — Ixorah can be working for you within weeks.