Farmland LP
The Vision
Farmland LP is the largest US fund manager focused solely on organic and regenerative farmland, managing a $350M+ portfolio. With a mandate to meaningfully scale acres and AUM, the team needed to manage materially more land, investors, and operational complexity without scaling headcount in lockstep. The chosen lever for this scale was autonomous AI agents—but agents only work on a foundation they can actually reason over.
The Solution
We built a four-layer data-centric stack where internal applications and AI agents act as “optional visitors” to a single, machine-readable source of truth.
- Data Warehouse (Microsoft Fabric): A medallion pipeline integrates six distinct source systems (accounting, field ops, Dataverse, SharePoint, etc.) into a centralized Gold star schema and DirectLake semantic model.
- Semantic Layer (Python + LinkML): A LinkML schema serves as the authoring surface for the enterprise domain model. A Python business-rule engine acts as the single implementation of every non-trivial logic rule—meaning DAX measures, read APIs, and agent tools all call the exact same logic.
- Ontology (OWL + SHACL): A formal OWL ontology expresses the complex hierarchy and temporal dynamics of farms, fields, funds, and lease transitions. SHACL shapes strictly validate incoming data at ingestion.
- Knowledge Graph (Graphiti + LightRAG): The ontology is materialized as a temporal knowledge graph for agent memory. LightRAG provides intelligent retrieval over unstructured institutional documents.
The Outcome
By building a robust semantic foundation, we deployed a team of autonomous agents—budget assistants, variance flaggers, and document synthesizers—that can safely answer cross-system questions in plain English because they draw from a reconciled, single source of truth.
The takeaway: The right order for enterprise AI is Data → Semantics → Ontology → Agents. Most organizations skip the middle two steps and wonder why their agents hallucinate. By building a rigorous semantic foundation, we unlocked the rare operational leverage required to scale AUM exponentially with existing staff.