TradeshiftがAmazon QuickでレガシーBIからエージェンティックAIへ移行Evolving from legacy BI to agentic AI at Tradeshift with Amazon Quick

Tradeshift, the company behind a widely used cloud network for business commerce, invoicing, and supply chain finance, has replaced its legacy business intelligence stack with an agentic AI solution built on Amazon Quick and Amazon Bedrock. The change matters because it reflects a broader shift in enterprise analytics: moving away from static dashboards that require trained specialists to interpret them, and toward systems that let business users ask questions in natural language and receive answers with far less manual effort.

At the center of the migration is the concept of "agentic" analytics. Where a conventional BI tool renders predefined reports and requires an analyst to build queries, an AI agent is designed to plan a sequence of steps, retrieve data from underlying sources, call tools, and iterate toward an answer with limited human guidance. According to the account, this allows a business user to pose a question in plain English and let the system decide how to assemble the relevant data, rather than navigating menus or writing SQL. The stated result is faster, more self-service access to insight and quicker decision-making.

The technical foundation appears to rest on two AWS services. Amazon Bedrock is a managed service that provides access to a range of foundation models from multiple providers through a single interface, along with features for retrieval-augmented generation, agents, and guardrails. It is commonly used as the model layer for generative AI applications so that organizations do not have to host or fine-tune large models themselves. Amazon Quick, part of Amazon's push to bring generative and agentic capabilities into its analytics portfolio, extends the QuickSight lineage of cloud BI with natural-language querying and agent-style workflows. Together, the two are positioned to translate a user's request into governed queries against curated data and to return summarized findings.

A migration of this kind typically involves more than swapping a query interface. Legacy BI environments usually carry accumulated data models, access controls, and report definitions that must be reconciled with a new platform. For natural-language querying to be reliable, the underlying data generally needs a well-defined semantic layer so that terms used by the business map consistently to the correct tables and metrics. Without that grounding, generative systems are prone to producing plausible but incorrect answers, so governance, data quality, and clear definitions remain prerequisites rather than afterthoughts. The write-up frames the project as a modernization effort, which suggests attention to these foundations, though the specific implementation details are best confirmed against the original source.

The move also fits an industry-wide trend. Major analytics and productivity vendors have been embedding conversational and agentic features into their products, including Microsoft's Copilot across Power BI and its broader suite, Google's generative capabilities in Looker and its cloud data tools, and Salesforce's Tableau and Agentforce efforts. AWS has similarly integrated its Amazon Q assistant into analytics and developer tooling. In that context, Tradeshift's decision to consolidate on Amazon Quick and Bedrock is one example of a company betting that natural-language, agent-driven analytics can lower the barrier to data use across non-technical teams. As a customer story published on an AWS channel, the narrative is likely to emphasize the benefits realized on that platform, so readers should weigh it as a vendor-associated case study.

For organizations weighing a similar path, several considerations recur. Agentic systems can reduce reliance on scarce analyst time and shorten the loop between question and answer, but they introduce new needs around monitoring model outputs, controlling costs of foundation-model calls, and ensuring that sensitive data is handled within appropriate permissions. Establishing clear metrics, human review for high-stakes decisions, and traceability for how an answer was derived are commonly cited safeguards. The value of such a platform ultimately depends on whether users trust the results enough to act on them.

Overall, Tradeshift's reported transition illustrates how enterprises are reframing BI as an interactive, language-driven capability rather than a library of fixed reports. Whether measured gains in speed and adoption prove durable will depend on data governance and user trust, but the direction of travel toward agentic, self-service analytics appears consistent across much of the industry.