Turns shopkeepers' chat voice notes into draft orders for FMCG distributors
TL;DR Spokenlist listens to the voice notes kirana owners send their distributor ("two cartons of the 500 ml oil, one box of the small biscuit") and turns them into draft orders in the distributor's billing software for a clerk to confirm. Distributors pay a monthly fee. It saves typing and cuts wrong deliveries.
India's small shops reorder from local FMCG distributors, and much of that ordering now happens on chat apps, often as voice notes in Hindi or a regional language mixed with brand names. There's a reason for that habit. Literacy among people aged seven and over was 80.9% in 2023–24, and 77% in rural areas[1], and even owners who read just fine find dictating a list faster than typing it.
At the distributor's end a clerk replays each note, works out which of several pack sizes was meant and types the order into billing software. Mistakes mean wrong deliveries, returns and arguments. I found no published figure on how many orders arrive this way or how often they go wrong, so treat the scale as something you measure.
The seed pitched conversational voice shopping for consumers and I reframed it. Consumer voice and chat shopping already has big players - one big grocery app sells end to end over chat through a 2022 deal with the chat platform's owner[2], and conversational-AI firms build voice and chat agents for food and quick-commerce apps[3]. A new consumer app would struggle. The unserved spot is the trade order between shop and distributor.
Spokenlist sits on the distributor's business chat number. When a voice note arrives, it transcribes the speech, matches each item to the distributor's own product list (brand, variant, pack size), and builds a draft order. Anything uncertain is flagged in colour. A clerk confirms with one tap, the order goes into the billing software, and the shopkeeper gets a written summary back to check. Over time it learns each shop's usual items and slang.
The distributor pays and the shop doesn't. A typical buyer is an FMCG distributor serving a few hundred kirana outlets with two to five order clerks. Pricing would be a monthly fee by order volume, set below the cost of one clerk. Brand companies that want cleaner secondary sales data are a later buyer.
Indian-language speech models are now available cheaply. The government's Bhashini platform, launched in July 2022, offers more than 300 pre-trained language models through open APIs[4]. The main chat platform's per-message pricing, in force since July 2025, does not charge for messages a customer sends to a business[5], so receiving voice notes costs a distributor nothing.
There's no public figure for the number of FMCG distributors in India or for orders placed by voice note. The narrowest signal is behaviour - large consumer brands already pay conversational-AI firms for voice and chat ordering agents across many languages[3], which suggests the speech pieces work. Treat the distributor count as the first thing to establish in the field.
Conversational-AI firms build voice and chat agents for large brands, not for small distributors[3]. Grocery ordering over chat is for consumers[2]. Distributor apps and sales-force tools let shops order by tapping through a catalogue, but shop owners keep sending voice notes anyway (old habits and all that). I found no product focused on turning trade voice notes into orders, though my search was limited.
| Difficulty | Moderate (42/100) |
| Starting capital | ₹65–125 L |
| Time to launch | 6–9 months |
The hard part is accuracy on product names. "the small biscuit pack" or "the blue one" means nothing without each shop's history, and code-mixed speech recorded in a noisy shop is tough to transcribe. Connecting to the many billing tools distributors use is the second job.
No sector licence. Ordering software for trade needs no retail or telecom licence, but a few rules shape how you run it.
Building this outside India? The categories above usually have a local equivalent, so check with your own regulator.
Never let the model send an order without a clerk's tap, and keep it away from credit limits or payment decisions. A wrong pack size costs a delivery, a wrong credit call costs money.
Want to work out the AI layer for Spokenlist? Talk to Yash →
I like this as a bet for a founder with speech or NLP skills who's also willing to sit in dusty distributor offices. Do not build a consumer voice app. Prove first that drafts from real voice notes need no corrections on at least 80% of lines for three distributors' product lists.
Plan for roughly ₹65–125 L and 6–9 months to launch, much of it on getting product-name accuracy high enough for clerks to trust.
Only if it costs a distributor less than an order clerk and clerks rarely need to fix lines. Distributors run on thin margins, so prove the saving in pilots before you price it.
Yes, wherever small shops order from distributors over chat apps. You'd need speech models for local languages and should check your own data protection law first.
The free Spokenlist validation kit has the line-by-line capital breakdown, the effort scoring, the data and messaging checklist, a 14-day validation sprint with pass/fail targets, a 90-day launch plan, 10 interview questions for distributors, and 5 assumptions to test. Download the kit →
Thinking about building Spokenlist or something close to it? The make-or-break work is getting product-name accuracy high enough that clerks trust the drafts, and I can help you plan that test. If you need a hand, you know where to find me. Talk to Yash →
For illustration only. Names are fictional and ideas hypothetical; sourced facts are real but may be out of date. Nothing here is legal, financial or professional advice, and Yash Khatri is not personally liable for any use of it. Read the full disclaimer.