AI Enterprise SolutionsGPT & Claude

SalesOS — From Conversation to Checkout

Small online retailers lose sales because customers struggle to find suitable products and complete purchases without human assistance. SalesOS is an AI-powered sales assistant developed around a real electronics retailer, elen.az. It understands customer needs, searches real products, compares options, makes personalized recommendations, manages shopping carts and creates orders. Unlike traditional customer support bots, SalesOS guides customers through the purchasing journey, from their first question to order creation. Business owners monitor conversations and manage orders through a Control Panel. Our long-term vision is to make this technology accessible to any business through automated AI-powered onboarding.

GitHub

Preliminary score

66.5/ 100Promisingaverage of 2 judges

Value for the user18.0 / 25
A specific problem and a clear outcome from solving it.
Prototype and use of AI22.0 / 30
A working core scenario, and what AI actually contributes to the result.
Quality testing12.0 / 20
Tests, examples of failures, and comparison with the current approach.
Feasibility9.5 / 15
Data requirements, running costs and a clear next step.
Originality5.0 / 10
A meaningful difference from common solutions.

What the judges wrote

GPT-6 AstraPreliminary71/100
Strengths
Your real-store integration and source-backed product selection give Gemini a substantive role in a concrete retail workflow. Persistent carts, server-side price and stock validation, and merchant-managed orders move the prototype beyond recommendations, with clear boundaries around manual fulfillment and no payment processing. Your reported isolated-database integration test and explicit disclosure of the unverified live Telegram flow provide useful, bounded testing evidence.
To improve
Your next test should verify the complete Telegram-to-dashboard journey and document the add-to-cart bug’s cause, resolution and regression case. Measure order-completion rate or staff time against the previous recommendation-only workflow to establish business impact. Add an estimated Gemini and hosting cost per completed order and a bounded next deployment milestone before expanding to generic onboarding.
Claude Opus 5.5Preliminary62/100
Strengths
Your project is grounded in a real retailer, elen.az, and works on real product names, prices, stock and links rather than a toy catalogue. The scope moves from advice to action: Gemini tool calling drives cart add/update/remove with server-side price and stock validation, checkout capture and persisted orders in a Control Panel. You are transparent about what was reused, what was built during the hackathon and what is still roadmap, including the unverified live Telegram checkout. That honesty makes your claims more credible.
To improve
The track asks for a metric that justifies adoption, and you do not name one. Measure things like question-to-order conversion, recommendation accuracy on a set of real customer queries, or merchant time saved, and compare them with how elen.az handles these requests today. Make the testing section concrete: show the add-to-cart integration bug you found, how often the agent picks the wrong variant or fails grounding, and a scripted set of conversations with pass/fail results. Also add an estimate of Gemini Flash Lite cost per conversation or per order to support the SaaS pricing.

The team

Aliyev MeherremeliLead
developer

Disclosure

Required by the rules. Judges verify it against the repository.

Models: Google Gemini Flash Lite — conversational reasoning, product selection and tool calling.
Data: Real product data from elen.az, including names, descriptions, prices and product links. SQLite stores conversations, shopping carts and orders.
Components: Python backend, Gemini-powered agent pipeline, modular website parser, Telegram integration, SQLite database, shopping cart tools, order management and web-based Control Panel. The project adapts a previously developed Sales Agent architecture. Existing search, messaging and administrative components were reused. New hackathon work focuses on guided purchasing, persistent shopping carts, order creation, management interfaces and revised agent behavior. Automated onboarding remains a planned extension.
Built during the event: Confirmed

Quality testing

161 automated tests passed. An integration test verified the Gemini-powered workflow from product discovery to shopping cart and order creation using an isolated SQLite database. Testing covered product search, application behavior and the new commerce workflow. An integration issue was identified in the add-to-cart tool and investigated during development. The complete end-to-end Telegram order flow has not yet been independently verified. Compared with the previous recommendation-only approach, the new workflow supports persistent carts and order management.

Worth 20 of the 100 points.

Running it

SalesOS uses Python, Google Gemini, Telegram and SQLite. Configure the environment variables with the required API credentials, install dependencies and start the application following the repository README. Customers interact with the AI sales assistant via Telegram. The Control Panel provides conversation history, shopping cart and order management. The current integration uses elen.az product data.

Project ownership

The team keeps the rights to this project. Judges were given access for verification only.