Transforming retail store performance management with generative AI

Teracloud designed and implemented a conversational AI agent powered by Claude 3.5 Sonnet via Amazon Bedrock. This solution enables users to interact with store performance data using natural language, gaining real-time access to store, product, and KPI data, and receiving AI-driven recommendations (e.g., stock optimization, pricing adjustments). The architecture was built on a secure and scalable AWS environment, leveraging services like AWS Lambda, Amazon S3, and Amazon API Gateway to ensure flexibility, integration, and scalability without infrastructure management.

About the client

The company specializes in digital solutions for the retail sector, driving operational efficiency in physical stores through customized software and innovative technologies. With a growing presence in Latin America, it enables smarter retail operations across multiple industries.

Challenges

They needed an innovative solution that would allow its users to query store performance metrics in real time, identify KPI deviations, and receive automated, actionable recommendations. The key objective was to deploy an intelligent agent integrated with their internal systems, eliminating manual reporting and accelerating data-driven decisions.

Project Goals & Objectives

The primary goal of the project was to enhance decision-making and operational agility across their retail store network by leveraging generative AI capabilities within a secure and scalable AWS environment.

Specific objectives included:

Enable real-time performance monitoring of individual stores using a conversational interface.
Automate actionable recommendations based on historical and real-time KPIs to support sales, inventory, etc.
Integrate seamlessly with the company's internal APIs, allowing efficient access to structured store, product, and KPI data.
Lay the groundwork for future enhancements, including analytics dashboards, knowledge base integration, and Lakehouse data architecture.

Why AWS

The AWS services were selected for their flexibility, seamless integration with the company's existing APIs, and ability to support rapid prototyping and production-ready deployment:

Amazon Bedrock – Enabled the use of Claude 3.5 Sonnet to provide advanced generative AI capabilities for natural language queries related to store performance.
AWS Lambda – Provided serverless compute for backend logic, handling API interactions, response formatting, and prompt enrichment without managing infrastructure.
Amazon S3 – Used as a central storage layer for structured datasets, supporting future analytics and data modeling needs.
Amazon API Gateway – Facilitated secure, scalable exposure of endpoints that connect the conversational agent to the company's internal systems.

Why us

The client chose Teracloud because we have experience designing and implementing conversational AI agent solutions. This involved leveraging advanced generative AI capabilities (Claude 3.5 Sonnet via Amazon Bedrock) and building a secure, scalable AWS architecture with services like AWS Lambda, Amazon S3, and Amazon API Gateway, which were justified for their flexibility, seamless integration, and ability to support rapid, production-ready deployment.

Partner Solution

Real-time and historical access to store, product, and KPI data.
AI-powered recommendations, such as stock optimization or pricing adjustments.
A secure production environment within the company's AWS account.
A roadmap for enhancements, including a Lakehouse architecture using S3 and Athena, strategic dashboards, and a per-store RAG-based knowledge base for enriched contextual responses.

Technical Solution & Justification

The AWS services were selected for their flexibility, seamless integration with the company's existing APIs, and ability to support rapid prototyping and production-ready deployment:

Amazon Bedrock – Enabled the use of Claude 3.5 Sonnet to provide advanced generative AI capabilities for natural language queries related to store performance.
AWS Lambda – Provided serverless compute for backend logic, handling API interactions, response formatting, and prompt enrichment without managing infrastructure.
Amazon S3 – Used as a central storage layer for structured datasets, supporting future analytics and data modeling needs.
Amazon API Gateway – Facilitated secure, scalable exposure of endpoints that connect the conversational agent to the company's internal systems.

Solution Optimality & Alternative Approaches:

Real-Time API Integration vs. Static Reporting – Lambda functions directly query the company's internal APIs to provide live data, ensuring that store performance insights are always up to date.
Alternative considered –Scheduled report generation; rejected due to latency and lack of interactivity.
Data Storage with S3 vs. Traditional Database – While not yet a full Lakehouse, storing structured data in Amazon S3 lays the foundation for future analytical use cases without requiring database administration.

Alternative considered:

Amazon RDS – dismissed for the current scope as it adds complexity without immediate benefit.
Serverless Execution vs. Container-Based Deployment – Using Lambda reduces operational overhead and ensures scalability without provisioning or managing infrastructure.
Alternative considered – Amazon ECS or EKS; rejected for this phase due to higher configuration and maintenance effort.

By focusing on flexibility, real-time access, and serverless architecture, the solution supports the company's current goals while providing a strong foundation for future enhancements such as AI-powered analytics and personalized knowledge bases.

Results and Benefits

Following the deployment of the generative AI agent, the company's experienced tangible improvements in store-level operations and user engagement:

50% faster access to store performance KPIs through natural language queries.
70% reduction in time spent generating manual reports by commercial and operations teams.
Improved accuracy and consistency in responses, leading to more confident and timely decision-making.
Increased adoption of data-driven practices across field teams and store managers.

Next Steps

Following the deployment of the generative AI agent, the company's experienced tangible improvements in store-level operations and user engagement:

50% faster access to store performance KPIs through natural language queries.
70% reduction in time spent generating manual reports by commercial and operations teams.
Improved accuracy and consistency in responses, leading to more confident and timely decision-making.
Increased adoption of data-driven practices across field teams and store managers.

Unmet Goals & Lessons Learned

While the implementation delivered clear value, several goals remain in progress or emerged as new opportunities:

Unmet Goals:

Real-Time Visual Dashboards: Although users can retrieve insights via chat, an embedded dashboard with filters and visual indicators will enhance data exploration and monitoring.
Advanced Data Aggregation: The current agent interacts with real-time APIs only. Historical data aggregation via Athena or similar tools is planned for the next phase.
Store-Specific Knowledge Base (RAG): Tailoring the agent’s context with per-store documents is a key roadmap item for deeper personalization and long-term learning.

Lessons Learned:

End User Involvement Drives Adoption: Early engagement with commercial stakeholders helped refine prompt formats and align responses with field expectations.
API Design Affects Agent Responsiveness: Latency and pagination logic had to be carefully managed to ensure the AI agent could deliver fast, relevant answers.
Scalability Requires Planning: Even in a serverless model, considerations around API limits and concurrent invocations were essential to ensure consistent performance.

About Teracloud

Teracloud is an advanced consulting partner in the AWS Partner Network, specializing in AWS and generative AI technologies. Our mission is to deliver transformative solutions that drive business value and innovation. As an AWS Advanced Tier Services Partner, we leverage our deep expertise and industry knowledge to help clients like UbuntuLaw achieve their strategic goals.