Big-data streaming company Confluent Inc. is beefing up its Confluent Intelligence platform to help artificial intelligence agents dig up more accurate data-based insights.

Its new additions include Streaming Agents, which use the Agent2Agent protocol to connect with specialized external AI agents, and a Multivariate Anomaly Detection tool that can help companies to prevent major outages.

Confluent is the developer of the open-source Apache Kafka data streaming platform, which is used by companies to track data points such as sales, orders, trades and customer feedback in real time. With Confluent’s managed version of Kafka, companies can feed this information into data analytics tools in real time, the moment it’s created, to obtain fresh insights about the state of their business. It also offers a managed version of the Apache Flink data streaming platform.

Confluent Intelligence is a relatively new offering. Launched in October, it’s a fully managed, cloud-native service that aims to connect real-time Kafka and Flink data streams to AI systems. Organizations can use it to develop context-rich AI agents and applications that can access live data, making them much more capable than those that can only access stale, historical data.

The new Streaming Agents capability in preview from today is all about enhancing agent’s abilities and facilitating agent-to-agent collaboration. Confluent explained that most AI agents in operation today work in isolation, unable to talk to each other or share business context, limiting their usefulness.

By using Google LLC’s Agents2Agents protocol, Streaming Agents provides a way for companies to enable more collaborative agentic automation. The Streaming Agents continuously analyze information from sources including BigQuery, Databricks, Snowflake and Langchain, then feed the insights they uncover to third-party agents in enterprise platforms such as Salesforce and ServiceNow. Those agents can then take immediate action based on up-to-the-minute insights.

The company said the result is that organizations can build much smarter and more responsive agents by feeding them with fresh context, but the real advantage unlocked by Streaming Agents is communication. It also provides a way for companies to orchestrate communication between multiple third-party agents and reuse their outputs across other systems. Streaming Agents acts as the orchestrator, while Confluent provides the governance, security and observability for all agent interactions.

Confluent Head of AI Sean Falconer said that as AI agents become more pervasive in enterprise environments, it’s imperative that they’re able to work together. “Confluence Intelligence connects team’s AI investments and systems no matter where they’re built, so AI can automatically react to live data, take action, coordinate systems and escalate issues when needed,” he said.

Meanwhile, the Multivariate Anomaly Detection feature is meant to enhance application and systems monitoring. Traditional anomaly detection tools tend to analyze telemetry data and metrics in isolation, and many times they’re restricted to batch-based analysis on historical data, rather than live information. The inability to perform analysis in real time means they cannot be used proactively, while the lack of context makes them susceptible to noise, spikes and false positives.

Multivariate Anomaly Detection, available in early access, does things differently, analyzing related metrics from across systems to obtain more context about applications and the systems they rely on. By analyzing multiple metrics at once, teams can filter out more of the false positives work proactively to prevent outages and downtime, the company said.

The feature can also help teams to spot more complex issues. For instance, by looking at signals such as memory, latency and processing performance at the same time, they’ll be able to identify patterns that might be missed when only looking at individual metrics.

With these updates, Confluent is addressing a fundamental challenge: the need for real-time relevance and collaborative automation. Now agents can access and communicate with each other based on the freshest streaming data, it’s providing enterprises with a foundation for building more proactive and capable AI agents that react instantly to business change.

Photo: Confluent