Retail Insights Agent

Inspiration

We were inspired to build a smart retail assistant that goes beyond answering basic inventory questions. Our goal was to combine Salesforce product data with live market trends to help retailers make smarter decisions about what to procure and sell, bringing AI-driven intelligence to everyday retail operations.

What it does

External services scan social media and blog platforms to detect trending products and brands.
These trends are compared with the Salesforce product catalog to filter relevant items.
XGBoost models forecast sales and growth trends for the shortlisted products.
If predicted demand exceeds current inventory, the system recommends restocking. The Salesforce developers built and trained the agent in Salesforce and connected it to the external Python service using RESTful APIs. The result is an agent that not only answers queries about product and inventory but also provides AI-powered suggestions based on real-world trends.

How we built it

Our team included Salesforce developers and AI/ML developers. The project architecture includes: Programming Language: Python Framework: Flask (for building RESTful APIs) Database: Microsoft SQL Server (MSSQL) Machine Learning: XGBoost (for sales forecasting models) External Services: Scripts and tools that analyze social media platforms and blog content for real-time trend data

Challenges we ran into

Our biggest challenge was connecting to the external service API to retrieve the latest trending data. Ensuring smooth communication, handling data format inconsistencies, and making the integration reliable and real-time required multiple iterations and close coordination between the teams.

Accomplishments that we're proud of

Successfully integrated a real-time trend analysis service with Salesforce via a custom-built API.
Developed a working AI model using XGBoost to predict sales and guide procurement decisions.
Delivered a seamless agent experience capable of both answering product queries and offering intelligent, data-backed recommendations.
Achieved effective collaboration across cross-functional roles—AI/ML and Salesforce.

What we learned

This project helped us understand how to bridge the gap between AI/ML services and CRM platforms like Salesforce. We learned how real-time trend data and sales forecasting models can work together to drive better inventory planning and gained practical experience in integrating machine learning systems into business workflows.

What's next for Retail Insights Agent

We plan to enhance the recommendation logic by incorporating more data sources such as customer purchase history and more. We also aim to integrate alert-based notifications for low-stock, high-demand products and explore UI enhancements within Salesforce for a more intuitive user experience. Long-term, we envision scaling this solution across multiple business units with configurable trend filters and customizable forecasting parameters.