Elasticsearch Has Changed Dramatically
The Elasticsearch you deployed three years ago is not the same platform available today. Between native ML nodes, ONNX Runtime integration, vector search, and — for those who look beyond the Platinum tier — community plugins that bring edge AI to Basic and Gold licences, the gap between what is deployed and what is possible is enormous.
Vector Search: From Keyword to Semantic
Elasticsearch’s k-NN (k-Nearest Neighbour) vector search lets you find semantically similar documents even when they share no keywords. A search for “cardiac arrest” returns documents about “heart attack” without those words appearing. This requires generating embeddings for your documents — something the Elastic Edge AI plugin suite automates for Basic/Gold tier deployments.
Named Entity Recognition at Ingest Time
NER at ingest time automatically tags documents with extracted entities (people, organisations, locations, dates) as they are indexed. This enriches your data without a separate ETL step and enables powerful faceted search on entity types.
Data Lineage with OpenLineage
The elastic-lineage plugin (part of Elastic Edge AI) adds OpenLineage-compatible field-level data lineage to Elasticsearch — tracking exactly which source fields contributed to which indexed fields. For compliance (GDPR, SOX) and debugging complex ingest pipelines, this is transformative.
Agentic Workflows: elastic-agent-studio
The newest category of Elasticsearch capability is agentic action: not just retrieving data, but acting on it. The elastic-agent-studio plugin closes the loop from “find this data” to “take this action based on what you found” without leaving your Elasticsearch infrastructure.
Why 70% of Users Miss These Features
The core reason is licensing: many of the best AI features in Elastic’s stack require Platinum licences ($X,000s/month at scale). The Elastic Edge AI open-source project was built specifically to fill this gap, running ONNX-based AI inference using the JVM runtime that already ships with every Elasticsearch instance.