Confluent, a data streaming platform provider, has announced new features for Confluent Intelligence. The enhancements aim to help companies better connect AI systems and utilize real-time data for analysis.
The two main innovations are Streaming Agents with the Agent2Agent protocol (A2A) and multivariate anomaly detection.
Connecting AI Agents in Real Time
Companies are increasingly using AI agents to automate processes. However, these agents often operate in isolation across different systems. This limits insights to individual areas.
Streaming Agents addresses this gap. The feature connects AI agents via the Agent2Agent protocol and integrates them with real-time data streams. It also uses Anthropic’s Model Context Protocol (MCP).
Agents can continuously process data from platforms such as BigQuery, Snowflake, or Databricks and transfer insights to systems like ServiceNow or Salesforce. Each action is recorded in an immutable log, enabling traceability and repeatability.
Apache Kafka coordinates communication between agents. This allows teams to supply existing agents with current context and manage asynchronous processes. Central orchestration is intended to ensure governance, security, and transparency across all interactions.
Early Detection of Anomalies
The second innovation concerns the integrated machine learning function. Multivariate anomaly detection analyzes multiple metrics simultaneously instead of examining individual values in isolation.
Conventional systems often rely on simple averages and are prone to false alarms caused by single outliers. The new function continuously learns from real-time data and filters out noise. This enables teams to better identify complex relationships – for example, between CPU utilization, memory, and latency.
The anomaly detection automatically adapts to changing data and does not require manual model training. Significant deviations from the learned normal state are flagged.
Cross-Industry Applications
The features can be used in various areas, including personalized offers in retail, risk assessments in finance, care recommendations in healthcare, maintenance predictions in manufacturing, and fault detection in telecommunications networks.
According to an IDC forecast, around 40 percent of positions in large companies will require collaboration with AI agents by 2026.
A2A support in Streaming Agents is now available in Open Preview. Multivariate anomaly detection is available in Early Access.
Sean Falconer, Head of AI at Confluent, emphasizes the need to connect AI systems with current data so they can respond dynamically to changes.
Dr. Jakob Jung is Editor-in-Chief of Security Storage and Channel Germany. He has been working in IT journalism for more than 20 years. His career includes Computer Reseller News, Heise Resale, Informationweek, Techtarget (storage and data center) and ChannelBiz. He also freelances for numerous IT publications, including Computerwoche, Channelpartner, IT-Business, Storage-Insider and ZDnet. His main topics are channel, storage, security, data center, ERP and CRM.
Contact via Mail: jakob.jung@security-storage-und-channel-germany.de