Skip to main content
In this tutorial you’ll build an outbound assistant that follows up on loan applications in Hindi.
1

Start from a template

Go to AI Assistants and click NBFC Loan Follow-up. Set the Agent name, for example Neha, and replace the lender details in the instructions with your own.
2

Set the language

On the Voice tab:
  • Primary language: Hindi
  • Auto switch language: on, so the assistant follows callers who reply in English
3

Choose a Hindi voice

Select Sarvam (model Bulbul v3) and preview the voices. Pick one that sounds natural for your audience and keep the speed close to 1.0x.
4

Translate the instructions

On the Prompt tab, click Translate prompt. Orato rewrites the instructions in Devanagari with a natural, spoken tone. Placeholders like {customer_name} and your section structure stay as they are.Review the result, then update the First message in Hindi, for example:
5

Check the transcriber

On the Advanced tab, Orato picks a transcriber from your languages. To set it yourself, click Change and choose Sarvam AI → Saaras v4 Realtime.
6

Pick the AI model

On the Settings tab, set Role to Outbound and choose a caller ID. Try Anthropic Claude Haiku 4.5 first, then compare with Sarvam AI → Sarvam 105B.
7

Test with native speakers

Use Talk and ask colleagues who speak Hindi to try it. Listen for unnatural phrasing, wrong number formats, and how it handles Hinglish.

Tips

  • Ask the assistant to say amounts the way callers expect, for example in lakhs.
  • Keep compliance language (fees, interest, consent) in Guardrails, and review the translated version carefully.
  • See Multilingual assistants for other Indian languages.