Confluent Embeds AI Agents Into the Data Stream to Enable Real-Time Agentic AI

(Piotr Swat/Shutterstock)

Why are so many enterprise AI agents still stuck in pilot mode? Companies have invested heavily in building intelligent systems, but most never make it past the testing phase. The problem is not usually the models themselves. It is a challenging task of continuously feeding agents the right data and connecting them to the tools that let them take meaningful action.

With its new Streaming Agents capability for Apache Flink, Confluent aims to solve that gap. The company says it enables AI agents to tap into live data streams, make context-aware decisions, and trigger automated actions at scale. The goal is to move beyond stalled experiments and help businesses deploy AI agents that actually deliver value.

The leap from prototype to production remains one of the biggest barriers to agentic AI. IDC reports that organizations launched an average of 23 GenAI pilots between 2023 and 2024, but only three reached production. Just 62% of those even met expectations. Many of these projects lose momentum once teams encounter the complexity of real-world workflows. Without timely context or strong integration, agents often fall short. 

“While most enterprises are investing in agentic AI, their data architectures can’t support the autonomous decision-making capabilities these systems require,” said Stewart Bond, Vice President of Data Intelligence and Integration Software at IDC. “Organizations should prioritize agentic AI solutions that offer easy, secure integration and leverage real-time data for the essential context needed for intelligent action.”

(innni/Shutterstock)

With its new Streaming Agents capability for Apache Flink, Confluent wants to change that. Instead of building AI agents that sit on the sidelines, disconnected from real activity, Streaming Agents places them inside the stream of what’s happening. These agents don’t wait around for updates. They tap into continuous flows of real-time data, stay aware of events as they unfold, and respond with full context.

Bridging the divide between AI ambition and real results is exactly what Confluent is trying to address. “Agentic AI is on every organization’s roadmap. But most companies are stuck in prototype purgatory, falling behind as others race toward measurable outcomes,” said Shaun Clowes, Chief Product Officer at Confluent. The issue, he noted, is not agent intelligence but the lack of fresh business context.

“Even your smartest AI agents are flying blind if they don’t have fresh business context,” Clowes explained. Streaming Agents was built to simplify the messy work of connecting data and tools into something usable. By giving teams a real-time foundation, the platform aims to help organizations move past early experiments and actually deploy agents that drive meaningful change across the business.

That foundation includes more than just live event data. Streaming Agents are built to connect with external APIs, transactional systems, and business applications. That gives them the ability to bring in relevant context and push decisions out to real systems. Whether it’s updating a database, starting a workflow, or sending a message to a customer, these agents can take meaningful action where it counts.

Running agents directly inside Flink also has practical benefits. It keeps deployment, testing, and monitoring within the same system that teams are already using. That lowers friction, especially for engineering teams that don’t want to integrate yet another tool just to get agents into production.

Confluent points to use cases where Streaming Agents are not just theoretical, such as competitive pricing. In ecommerce, the ability to adjust prices in real time can directly impact revenue. Confluent explains that Streaming Agents can monitor prices across competitor sites and automatically update a retailer’s own listings to reflect the most competitive offer. There is no need for manual checks or delayed updates. Just fast and continuous adjustments that help win the sale.

(NicoElNino/Shutterstock)

Another example is how agents connect to external tools and systems. Using Model Context Protocol (MCP), Streaming Agents can choose the right tool for the situation, whether it is a database, an API, or a business application. Based on what is happening in the data stream, agents can trigger actions like writing to a system, updating a record, or sending a message, all without waiting for human input.

These examples highlight the shift Confluent is aiming for. These agents do not sit idle. They stay active inside the flow of business activity, aware of what is happening, and ready to act with the right context. 

While there is promise, rolling out Streaming Agents is likely to present challenges. Implementation depends on infrastructure readiness and integration with existing systems. Confluent’s approach brings together Flink, Kafka, and secure tool connections to support this setup. The broader question for enterprises is whether their environments can support agents that operate continuously and respond to real-time signals. Intelligence alone may not be sufficient. What matters is whether agents can access the right context as conditions continue to change around it. 

Related Items 

Ataccama Introduces AI Agent For Enhanced Data Management

AI Agent Claims 80% Reduction in Time to Complete Data Tasks

Google Pushes AI Agents Into Everyday Data Tasks