AI agents are transforming analytical databases in two fundamental ways: they are changing the workloads databases serve, and they are changing how databases themselves are built.
Agents on the Database are creating entirely new workload patterns. Instead of static dashboards and predefined analytics workflows, AI agents are generating ad-hoc queries, exploring data autonomously, and becoming a new interface to enterprise software. At the same time, customers increasingly demand that these systems run wherever their data lives across public clouds, sovereign clouds, AI-native cloud providers, and on-premises environments.
Agents in the Database are changing how databases are built. Coding agents work best with open-source systems where they can inspect the code, understand behavior, debug issues, customize and add new functionalities. As a result, architectural decisions that once optimized for human teams are being re-evaluated through the lens of agent productivity. Simpler, modular architectures are easier for agents to understand, modify, operate, and verify, giving open-source systems a significant advantage in the age of AI.
In this talk, we will share Firebolt's journey from a proprietary cloud data warehouse to an open-source analytical database designed for the agentic era. We will discuss why we chose to open-source Firebolt, how AI agents influenced our architectural redesign, and why we are moving from a federation of multi-tenant services toward a modular monolith architecture.