Prioritizing Agentic Use Cases Across Heterogeneous Lakehouse Platforms

Determine a systematic framework for prioritizing AI agent use cases within data lakehouse environments, specifying which data lifecycle tasks should be targeted first across heterogeneous lakehouse architectures and interfaces to guide safe and effective agent deployment.

Background

The paper proposes a programmable, API-first lakehouse design and demonstrates a proof-of-concept for agentic repair of data pipelines. While the authors showcase one concrete use case, they explicitly note uncertainty about how to prioritize agent-driven automation across diverse, heterogeneous platforms.

This uncertainty reflects the breadth of possible agentic tasks in lakehouses—spanning data access, pipeline execution, observability, and governance—combined with differences in platform interfaces and operational constraints. A principled prioritization would help decide which agentic workflows should be addressed first in production settings.