Smart Extract: from a chat screenshot to a courier order in seconds.
Smart Extract is an AI-powered SaaS I founded and built for online sellers in Kosovo and Albania. A seller uploads a screenshot of a customer conversation, the app reads the name, phone number, city and address, and the order is registered with the courier, without typing anything.
- 35,000+screenshots processed, April to September 2026
- 5-10 sper order, instead of 1-2 minutes by hand
- 4courier services integrated
- 99.9%of screenshots processed successfully
The problem
Many shops in Kosovo and Albania sell through social media and take orders in chat. Every order then has to be typed by hand into a courier's website: name, phone number, city, address, product. That takes one to two minutes per order, and every typo in a phone number or address turns into a delivery problem.
What I built
- Screenshot to order. Upload one screenshot or a batch. Google Gemini reads the conversation and extracts the customer's name, phone, country, city, address and order details into a structured order.
- Direct courier registration. Integrated with four courier services: Postman, Delex, Cheetah and Express Posta. The extracted order goes straight into the courier's system.
- Web and mobile. A web app for daily work, an Android app on Google Play (iOS coming soon) and a public website.
- A partner API, so other systems can use the extraction directly.
- Billing built in. Subscriptions and pay-as-you-go credits, with invoices generated in Albanian and English.
How it's built
The backend is a .NET 10 API with Clean Architecture and CQRS, on PostgreSQL 18 and Redis. The web apps are Next.js, React and Tailwind CSS; the mobile app is React Native with Expo. The API runs in Docker on Google Cloud Run, with CI/CD in GitHub Actions.
Two decisions worth mentioning:
- Screenshots are processed in memory and not stored. Customer conversations are personal data, so the system keeps only the extracted order, not the image.
- Background work runs on a separate .NET worker. Subscription renewals, invoices, analytics and emails run on a small dedicated server instead of always-on cloud instances, which keeps running costs low for a product priced for small shops.
Results
Between April and September 2026, Smart Extract processed 35,797 screenshots, and 99.9% of them were processed successfully. Monthly volume more than tripled from April to August, after I started posting short videos on TikTok, Instagram, YouTube, LinkedIn and Facebook. Registering an order takes 5-10 seconds instead of 1-2 minutes, and sellers no longer retype addresses and phone numbers. Here is what some of them say:
“Using this application has greatly simplified our work of registering orders from social media to courier services. It has helped us register orders faster and with much more accuracy. Errors have also been significantly reduced. We recommend it.”
“Since we started using Smart Extract, order registration has become much easier and faster. From a single conversation screenshot, customer data is prepared automatically without needing to type anything manually. It has helped us save time, reduce errors, and manage our daily work more easily.”
“It has simplified about 80% of our order registration process, converting data directly from photos into ready-to-use information for shipping and management. The staff is excellent and always ready to respond whenever we need help.”
“Smart Extract has made our work so much easier and removed the stress of order registration. Great work you've done, respect.”
Built with Amira Hoxha, who made the website and the app's UI foundation.
Need something like this for your business?
If your team copies data from messages, photos or documents into another system by hand, that work can usually be automated. Tell me what it looks like today.