Several data-platform announcements are addressing a shared challenge for enterprise AI: making information available to applications and agents without detaching it from the systems, controls and formats organizations already use. Updates from Airbyte, Denodo, Hitachi Vantara and ClickHouse span semantic search, hybrid-data access, analytics on object storage and a broader database platform.
The approaches differ. Some aim to bring data from multiple sources into an AI assistant’s reach; others put query or application capabilities closer to stored data. Together, the announcements illustrate how vendors are positioning integration and data access as components of AI infrastructure, rather than treating model access as a standalone problem.
Airbyte connects search with structured data
Airbyte announced semantic search in its context layer, designed to help AI agents locate relevant information in customer conversations, documents and support tickets. The company says the feature can combine those findings with structured data in Snowflake and BigQuery, allowing users to ask business questions in natural language.
The capabilities are available through the Airbyte Model Context Protocol (MCP) server. Users can make requests from Claude, ChatGPT and other MCP-compatible assistants. The announcement describes a way to bring information from unstructured sources into the same question-and-answer workflow as data in the two named analytics platforms; it does not specify additional supported sources or publish accuracy benchmarks.
For organizations building agent workflows, the distinction matters: locating relevant text and combining it with structured records are separate data-access tasks. Airbyte’s update is aimed at addressing both through an assistant interface, although the announcement does not detail deployment requirements or the underlying search architecture.
Hybrid access and analytics close to stored data
Denodo signed a multi-year strategic collaboration agreement with Amazon Web Services focused on helping enterprises develop and scale agentic AI using real-time, governed data across hybrid and multicloud environments. The agreement builds on the companies’ existing relationship and represents a new phase of investment to accelerate adoption of the Denodo Platform on AWS worldwide. The announcement does not provide financial terms or a schedule for specific product changes.
Hitachi Vantara, meanwhile, expanded its VSP One platform with capabilities to analyse supported data directly within its object storage. The company says the approach can reduce the need to move or copy data into separate systems. Supported open formats include Apache Iceberg, Parquet, JSON and CSV. Hitachi reports up to five times faster analytical queries using 80% fewer CPUs; those performance figures are the company’s claims, and the announcement does not specify the test conditions.
The update also adds a four-node starting configuration and new Hitachi iQ Studio capabilities intended to connect AI applications and agents to governed enterprise data. Keeping analytics near object data can reduce reliance on a separate copy for supported workloads, while the smaller starting configuration gives customers another deployment option. The extent of the benefit will depend on workload and deployment details not included in the announcement.
Database and application-platform changes
ClickHouse acquired Xata, which the announcement describes as an agent-native Postgres company. ClickHouse says the acquisition will give developers and AI agents access to isolated, production-like data environments on demand. The company also outlined a wider platform that brings together transactional data, observability, security telemetry and AI engineering with its analytics engine.
That portfolio includes ClickHouse Managed Postgres, ClickHouse Observability and ClickHouse Security, with the latter two built on technology from ClickHouse’s September acquisition of RunReveal. Langfuse, the open-source AI engineering platform acquired in January, adds an AI Gateway intended to provide visibility and control over model use in applications. These are platform and product descriptions from ClickHouse; the announcement does not provide comparative performance or integration benchmarks.
Elsewhere, MongoDB said MongoDB Atlas for Government achieved FedRAMP Class D certification and added Microsoft Azure Government support, in addition to existing availability on Amazon Web Services and Google Cloud Platform. The update expands the deployment options the company identifies for that regulated offering.
These announcements reflect different integration priorities: connecting assistants to mixed data, supporting governed access across environments, querying data where it resides, and consolidating application workloads. They do not establish a single architecture for enterprise AI. Organizations evaluating the offerings will still need to assess supported data sources and formats, governance requirements, deployment models and workload-specific performance.