If you watch how a small business owner in India communicates with their customers and suppliers on any given morning, you'll notice something quickly: almost nobody is typing.
They're speaking. Thirty-second voice notes, sometimes longer. Orders placed by voice. Delivery confirmations by voice. Payment disputes, complaints, and pleasantries — all voice. Typing is for the message you need to be precise about. Everything else is voice, because voice is faster, more natural, and doesn't require you to put down what you're holding to hunt for letters on a screen.
Voice notes have won. The question is whether your tools have caught up.
The problem with voice in every other tool
Most business software treats voice as an afterthought. You can record a voice note in WhatsApp and attach it to a customer record in some CRMs, but the voice note is just a file. Nobody reads the file. Nobody summarises it. Nobody knows what's in it unless someone listens to it — again.
So the information in voice notes is practically invisible to any system that tries to help you organise your business. The customer said they'd pay by Friday? That's in the voice note. The delivery changed from Site A to Site B? Also in the voice note. Three weeks later, when you're trying to figure out why the payment never came, you have no idea where to start.
This isn't a small gap. If the majority of your business communication happens by voice and your tools can't read voice, your tools are essentially blind to most of what's happening in your business.
What makes voice hard to understand
Transcribing a voice note in a controlled environment — English, single speaker, quiet background — is a solved problem. But that's not the environment. The environment is:
- Hindi and Punjabi mixed with English, sometimes mid-sentence
- Regional accents and local business vocabulary ("nag", "patti", "diggi") that don't appear in training data for most models
- Background noise — market sounds, vehicles, other conversations
- Short clips under ten seconds with no context
- The same speaker switching between three languages and a mix in the same voice note
Most transcription systems fail or produce near-useless output in this environment. We spent weeks working on this before HERE's first pilot, specifically because getting voice right wasn't optional. If we couldn't understand what the merchant was actually saying, the whole premise fell apart.
What HERE does differently
HERE uses Sarvam — an AI model built specifically for Indian languages and code-mixed speech. It handles Punjabi, Hindi, English, and natural mixes between them. It understands that when someone says "Rajinder ke liye kal 15 sheets bhej dena" they mean an order, not a general note. It knows that "patti" is a material unit, not a word to be puzzled over.
After transcription, HERE extracts the structured facts — who, what, how many, when, how much — and stores them as something you can query later. Not a voice note to listen to again. An answer you can get in three seconds.
This matters more than it might sound. When a merchant asks "Rajinder ne kal kya bola tha?" the answer isn't another audio file. The answer is: "15 sheets, delivery tomorrow, same site, payment next week." That's the version of memory that's actually useful.
Why this is a bigger deal than it sounds
The framing of "voice interface" usually comes with a lot of science-fiction overhead — voice-activated assistants, smart speakers, ambient computing. That's not what we're talking about.
We're talking about a specific, practical problem: the most common form of business communication in a large and underserved market produces information that existing tools cannot read. HERE reads it.
That's the whole thing. It's not glamorous. But it's the thing that makes HERE actually useful rather than theoretically interesting.
If your business runs on voice notes, any tool that can't understand voice notes understands maybe 30% of your business. HERE understands the rest.
We're in private pilot now with merchants across India and Canada. If you run a business where voice is the primary channel — and most small businesses in India do — we'd like to hear from you.