Confluent, Inc. (Nasdaq: CFLT), a pioneer in the field of data streaming, has introduced Streaming Agents, a new feature within Confluent Cloud for Apache Flink®. This tool is designed to simplify the creation and scaling of AI agents capable of monitoring, analyzing, and acting on real-time data.
The goal is to eliminate the main barriers organizations face when implementing large-scale artificial intelligence. Streaming Agents unify data processing with AI workflows, providing secure, direct connections to all parts of a company—from large language models (LLMs) to integration systems, tools, and enterprise platforms.
This promotes the adoption of agent-enabled AI, leading to more efficient processes, faster return on investment, and the potential to generate new business models and growth opportunities.
“Agent-enabled AI is part of every organization’s roadmap. However, most companies are stuck in the prototyping stage, falling behind while others move toward measurable results,” says Shaun Clowes, Confluent’s Head of Product. “Even the smartest AI agents act blindly without up-to-date business context. Streaming Agents simplifies the complex task of integrating tools and data that create real intelligence, providing organizations with a solid foundation to deploy AI agents that drive meaningful change across the enterprise,” he adds.
According to an IDC study, although organizations conducted an average of 23 generative AI proof-of-concept tests between 2023 and 2024, only three reached the production phase. Of those, just 62% met expectations. Agents are as powerful as the tools and data they can access, but current workflows are incredibly complex and costly, preventing companies from fully leveraging agent-enabled AI. While existing AI frameworks make it easier to start with agents, many teams struggle to integrate real-time data into their agent initiatives, resulting in hallucinations and unreliable responses.
“Although most companies are investing in agent-enabled AI, their data architectures can’t support the autonomous decision-making capabilities these systems require,” notes Stewart Bond, IDC’s Vice President of Data Integration Software. “Organizations need to prioritize agent-enabled AI solutions that offer easy, secure integration and leverage real-time data to provide the critical context needed for intelligent action.”
Create and scale AI agents in real time with Streaming Agents
Streaming Agents bring agent-enabled AI directly into streaming processes to help teams create, deploy, and coordinate event-driven agents using Apache Kafka® and Apache Flink®. By unifying data processing and AI reasoning, agents gain access to updated contextual data from real-time sources, allowing them to adapt swiftly and communicate with other agents and systems as conditions change. These agents are always active, working on behalf of the company, dynamically processing high-volume data streams, and instantly responding to real-time signals with contextually sensitive reasoning akin to human operators.
For example, Streaming Agents can perform competitive pricing by continuously monitoring prices on e-commerce sites and automatically updating a retailer’s website product rates to offer the most competitive deals to customers.
Key features of Streaming Agents include:
Tool Calls for Context-Aware Automation: Invoking tools via the Model Context Protocol (MCP) allows agents to select appropriate external tools, such as databases, SaaS, or APIs, to carry out meaningful actions. Tool calls consider ongoing business activities and other systems and agents’ actions.
Secure Integration Connections: Connect securely to models, vector databases, and MCP directly through Flink. These connections also protect sensitive credentials, promote reuse by sharing connections across multiple tables, models, and functions, and centralize management for large-scale deployments.
External Tables and Search to Enhance AI Accuracy: Ensure streaming data is enriched with external data sources—like relational databases and REST APIs—for the most up-to-date, comprehensive data view. This improves decision-making accuracy, vector search, and retrieval-augmented generation (RAG) applications, while reducing cost and complexity with Flink SQL and leveraging Confluent Cloud’s security and networking features.
Replay Capability for Iteration and Security: Agents can be developed and tested using real data without impacting live systems, enabling silent launches, A/B testing, and faster iterations.