Evolving from legacy BI to agentic AI at Tradeshift with Amazon QuickSight
Tradeshift replaced its legacy in-house BI tool with Amazon Quick, achieving a 40% reduction in total cost of ownership and eliminating the need for manual engineering maintenance. Query performance improved dramatically, with response times dropping from 45-90 seconds to under three seconds for datasets ranging from one million to one hundred million records. The implementation introduced agentic AI capabilities, including natural language Q&A via the AP Auditor chat agent and automated workflo
Analysis
TL;DR
- Tradeshift replaced its legacy in-house BI tool with Amazon Quick, achieving a 40% reduction in total cost of ownership and eliminating the need for manual engineering maintenance.
- Query performance improved dramatically, with response times dropping from 45-90 seconds to under three seconds for datasets ranging from one million to one hundred million records.
- The implementation introduced agentic AI capabilities, including natural language Q&A via the AP Auditor chat agent and automated workflows through Amazon Quick Flows and Research.
- The solution was structured as a tiered embedded analytics product, democratizing data access for both internal finance teams and external buyers/sellers across 70+ countries.
- Security and multi-tenancy were ensured through a four-layer architecture involving Okta SSO, custom namespaces, signed URLs, and row-level security rules.
Why It Matters
This case study demonstrates the practical viability of transitioning from rigid, legacy business intelligence systems to agentic AI-driven platforms, highlighting significant gains in operational efficiency and user experience. It provides a concrete blueprint for enterprises looking to embed advanced analytics directly into their products, turning data insights into a revenue-generating feature rather than just an internal utility. The success story underscores how agentic AI can resolve bottlenecks in manual reporting and enable self-service analytics for non-technical users at scale.
Technical Details
- Performance Metrics: The new system processes 1-100 million transaction records with sub-three-second latency, a 30x improvement over the legacy tool's 45-90 second response times.
- Agentic Features: Utilization of Amazon Quick’s chat agent for natural language querying, Flows for coding-free automation of multi-step workflows, and Research for generating long-form analytical reports.
- Architecture Layers: The deployment consists of 16 embedded Amazon Quick Sight BI dashboards via secure iFrames, a conversational AI layer for NLQ, and a tiered access model (Standard vs. Premium) offering varying levels of data autonomy and customization.
- Security Framework: Implemented a robust four-layer security model including Okta single sign-on, Amazon Quick custom namespaces for tenant isolation, signed URLs for session-bound embedding, and approximately 14,000 row-level security (RLS) rules to filter data by user context.
- Implementation Timeline: The project followed a phased approach starting with a proof of concept in early 2024, moving to an MVP by March 2025, and achieving full organizational adoption (98%) by August 2025.
Industry Insight
Companies should evaluate agentic AI not just for automation but as a core component of product strategy, enabling the creation of self-service analytics tools that reduce dependency on specialized data engineering teams. The shift from static dashboards to conversational, agentic interfaces allows for deeper user engagement and democratizes data access, particularly in complex B2B marketplaces where diverse user groups require tailored insights. Furthermore, leveraging managed cloud services for embedded analytics can significantly lower total cost of ownership while enhancing scalability and security, making it a compelling alternative to maintaining proprietary, legacy BI infrastructure.
Disclaimer: The above content is generated by AI and is for reference only.