CodeInventory with live stock levels
Every product carries its current stock count, updated as sales and purchases are recorded — the shop's real position, not a guess.
Case study — AI inventory app for kiryana stores
Dukaan AI
An AI shopkeeper's assistant that watches stock, forecasts demand and warns before anything runs out.
AI ProjectIn productionNext.jsFirebaseFlutter
Overview
About this project
Dukaan AI is an inventory and sales assistant built for kiryana and retail shopkeepers who run their whole business from memory and a paper register. It tracks stock, records sales, forecasts what will sell next, and raises reorder alerts before shelves empty — with an AI assistant the shopkeeper can simply ask, in Urdu or English, 'what is about to finish?' A web dashboard is paired with an offline-first Flutter app, because shop internet is not always there when the customer is.
The challenge
The problem it solves
A kiryana store carries hundreds of fast-moving items, and the typical owner keeps stock levels in his head. It mostly works — until a staple runs out unnoticed, a customer leaves empty-handed, and the same mistake repeats next week. Paper registers record what was sold but never warn about what is about to finish, and no small shop can afford inventory software built for supermarkets. The challenge was to build stock awareness that fits a one-person shop: quick to update, forgiving of missed entries, and smart enough to turn sales history into a reorder warning in plain Urdu.
What we built
Features that do the work
Low-stock and reorder alerts
When an item falls below its threshold, Dukaan AI flags it for reorder before the shelf empties — the alert the paper register could never give.
Sales recording and demand forecast
Daily sales build a history the system turns into a simple forecast of what will move next, so purchasing follows demand instead of habit.
AI assistant chat in Urdu and English
The shopkeeper asks questions in his own language — stock checks, today's sales, what to reorder — and gets direct answers from the shop's own data.
Private account for every shop
Each shop signs in with its own account; inventory, sales and forecasts are stored per user and visible to no one else (production upgrade, October 2026).
Offline-first Flutter app with sync
The mobile app keeps working without internet — sales and stock changes are stored on the phone and sync to the cloud when a connection returns.
Technology stack
Next.jsFirebase Auth & FirestoreFlutter appAI chat
Project journey
How Dukaan AI got here
- October 2026
Concept & design
The concept started from the counter, not the office: a shopkeeper with thirty seconds between customers. Every screen was designed for one-thumb, one-glance use, in Urdu first.
- October 2026
Build
The web dashboard, inventory engine, forecasting and AI chat were built, together with the Flutter app sharing the same data.
- October 2026
Quality pass
The full flow was tested the way a shop actually runs: offline sales during the day, sync in the evening, and alerts checked against real stock counts.
- 21 October 2026
Production upgrade
Dukaan AI was upgraded from a demo to a production architecture: real user accounts, per-shop private data with locked-down access rules, and a server-side AI engine replacing canned responses.
Current status
Where it stands today
Production-ready build complete: accounts, private per-shop data, the server-side AI engine and the offline Flutter app are all in place. Deployment is in progress.
Next step: Go-live on its own subdomain once the production hosting and Firebase project are connected, followed by the Android APK build through the cloud build pipeline.
Client words
What the client says
“I keep my whole shop in my head, and mostly that worked — until stock ran out without me noticing. The app warns me before things finish, and it speaks to me in Urdu, so I did not need anyone's grandson to set it up.”
Want results like Dukaan AI?
Tell us about your business on WhatsApp — we will map the fastest route from where you are to a project like this one, with a fixed quote before anything starts.