New platform capabilities add Query as the fifth pillar of data streaming, enabling AI agents and applications to access continuously updated, governed business data.
Confluent, an IBM company, has introduced its Streamhouse architecture, designed to help enterprises provide AI agents, applications and analytics with a continuously updated view of business operations. The announcement expands Confluent’s Data Streaming Platform (DSP) with Query as its fifth pillar, alongside Stream, Connect, Process and Govern, creating a unified foundation to capture, process, govern and serve real-time business data.
The launch comes as enterprises increasingly deploy AI agents to make decisions and execute tasks across business functions. However, fragmented data environments and systems that update at different intervals can leave AI models working with outdated or conflicting information. Streamhouse aims to address this gap by maintaining a governed, current view of business activity that can be accessed at machine speed.
A key addition is Lightning Tables, now generally available on AWS through Confluent Cloud. The read-only serving layer materialises Apache Kafka data into continuously updated tables, enabling applications to access live business information through SQL and APIs without requiring separate databases, caches or additional ETL pipelines.
“An agent can only make a useful decision if it can see what is true across the business when it acts.”
— Greg Taylor, SVP & General Manager, APAC, Confluent
In Confluent’s internal testing, Lightning Tables delivered approximately 100 milliseconds from data creation to serving, query latency of 10–100 milliseconds and throughput exceeding 1,000 queries per second. Actual performance may vary depending on configuration and workload.
The new Query pillar also brings together Confluent Tableflow, which makes streaming data available to data warehouses and lakehouses in open table formats, and the Real-Time Context Engine, which provides AI agents with current business context through the Model Context Protocol (MCP).
The platform combines Apache Kafka and the Kora engine for event streaming, more than 150 managed and custom connectors, Apache Flink for real-time data processing, and governance capabilities covering data quality, lineage, encryption, schemas and access controls. It is designed to work alongside existing data warehouses and lakehouses rather than replace them.
Confluent cited Accenture as an enterprise already using a Streamhouse architecture to connect data across its global operations. According to Thomas Mahaffey, SAI Chief Data Strategist at Accenture, the approach has reduced the organisation’s time-to-insight from months to days and significantly shortened reporting cycles. The architecture supports data operations across more than 120 countries and serves the needs of its 800,000 employees.
Shaun Clowes, General Manager, Confluent and Data, IBM, said the architecture gives agents and applications a live, governed view of business activity, helping organisations respond faster to fraud signals, customer requirements and operational changes.
The Streamhouse approach rests on three principles: keeping business context current, supporting mission-critical production requirements and accessing decentralised data where it resides. By connecting operational data with analytics and AI systems, Confluent is positioning real-time data infrastructure as a foundation for enterprise AI adoption.
