Extract
Not specified
Business Operations
Amazon Bedrock, Amazon S3, Amazon RDS PostgreSQL, Amazon DynamoDB, AWS Lambda, Amazon API Gateway, AWS Cognito, Amazon Verified Permissions, Amazon SageMaker, Amazon
Extract
Not specified
Business Operations
Amazon Bedrock, Amazon S3, Amazon RDS PostgreSQL, Amazon DynamoDB, AWS Lambda, Amazon API Gateway, AWS Cognito, Amazon Verified Permissions, Amazon SageMaker, Amazon
Extract, a workforce optimization and payments company, needed to deliver intelligent automation for dry cleaning businesses using its tailored operational toolset. Users needed to ask natural language questions, manage documents and photos, and get real-time operational insights from the XSYS system, all with fast responses and strict access controls. Avahi built an Amazon Bedrock Agent that combines retrieval augmented generation, document management, and Al-driven analytics and visualization in a multi-tenant, role-based platform. The solution achieved sub-5-second query performance while supporting up to 500 concurrent users with secure data isolation.
Extract specializes in workforce optimization and payment solutions and builds industryspecific operational tooling. For dry cleaning businesses, Extract delivers capabilities that span daily operations, documentation, and performance insights, with a focus on automation and better decision-making.
Dry cleaning operators often need quick answers across operations, staffing, payments, and customer workflows, but the information is fragmented across operational systems, documents, photos, and spreadsheets. Extract needed an Al-powered system that could unify these sources and let users ask questions naturally, then receive accurate, context-aware responses.
Beyond Q&A, the platform had to manage multiple file types, including PDFs, Word documents, spreadsheets, and images, then use that content for analysis and recommendations. It also needed to integrate directly with XSYS for real-time and historical operational data, and return outputs that fit business workflows, including tables and visualizations like bar charts, line graphs, and pie charts.
Because Extract serves many businesses, the system needed a secure multi-tenant architecture with strict data isolation and role-based access across Admin, Owner, and Staff roles. Performance was critical, with a requirement to support up to 500 concurrent users and deliver sub-5-second query response times
AWS provided the managed services needed to build a secure, scalable agentic platform without sacrificing performance. Amazon Bedrock enabled natural language understanding and agent behavior, while serverless components supported elastic scaling to meet concurrent usage demands. AWS also offered built-in identity, access control, monitoring, and durable storage services that are well-suited for multi-tenant SaaS architectures.
With AWS-native security services, Extract could enforce role-based access policies consistently across tenants and data types, and monitor performance to ensure the system met latency requirements.
Extract chose Avahi for the ability to deliver an end-to-end, production-oriented agent architecture that blends natural language automation with operational integrations and governance. Avahi designed the platform to work across structured and unstructured data, integrate with XSYS for live operational context, and generate business-ready outputs like tables and charts. Avahi also implemented secure multi-tenancy and granular permissions aligned to Admin, Owner, and Staff roles, while meeting strict performance targets under concurrent load.
Avahi built a comprehensive Amazon Bedrock Agent that serves as a unified operational interface for dry cleaning businesses. Users submit natural language questions through a responsive UI, and the agent returns context-aware answers with summarization, pattern recognition, recommendations, and analysis. The system combines operational data from XSYS with customer-uploaded files to provide grounded responses.
To support document and file understanding, Avahi implemented a retrieval augmented generation pipeline backed by a vector database. An ETL layer ingests documents, photos, and spreadsheets, transforms and normalizes content, and performs vectorization so the agent can retrieve relevant passages and data points at query time. Amazon S3 stores the raw files, DynamoDB stores semi-structured metadata and supporting records, and Amazon RDS PostgreSQL stores structured operational datasets used for analytics and reporting.
For business intelligence workflows, the solution generates structured outputs, including tables and Al-driven visualizations such as bar charts, line graphs, and pie charts. These are produced from the combined dataset (XSYS plus uploaded artifacts) and presented in the frontend experience for quick operational decision-making.
The platform integrates with XSYS via APIs to access real-time and historical operational data. Amazon API Gateway exposes secure REST endpoints, and AWS Lambda orchestrates the backend agent workflow, including retrieval, policy checks, transformation, and response assembly. Authentication is handled through AWS Cognito, and authorization is enforced with Amazon Verified Permissions to support Admin, Owner, and Staff roles with tenant-aware access policies. CloudWatch monitors system health and performance, enabling ongoing tracking of latency and throughput. The design supports multi-tenancy with data isolation and autoscaling to meet concurrent usage requirements.
Extract delivered a secure, multi-tenant Al agent that unifies operational data and business documents into a single natural language interface for dry cleaning businesses. The platform supports fast, actionable insights, including visualization and analytics outputs, while integrating with XSYS for live operational context. With role-based access controls and tenant isolation, the system improves operational efficiency at scale without compromising security.
Let’s explore your high-impact AI opportunities together in a complimentary session
Let’s explore your high-impact AI opportunities together in a complimentary session
Let’s explore your high-impact AI opportunities together in a complimentary half-day session
