Shutterstock
All this interest in AI agents is pushing data infrastructure vendors to rebuild their platforms to process more autonomous, event-driven workloads. Getting real-time context around your streams is emerging as a key requirement—a capability that most batch-based systems and conventional data lakes struggle to support.
Confluent, long known for its streaming data backbone built on Kafka, is positioning its latest updates as a response to that shift. At its Current 2025 user conference this week, the company announced a group of changes meant to bring streaming infrastructure closer to the world of AI-native development.
Confluent Intelligence was the standout feature. It is a controlled stack designed to enable teams to develop and manage AI agents utilizing real-time data. The company also launched a Private Cloud offering targeted at regulated industry players and expanded Tableflow to include Unity Catalog and Microsoft Azure, enabling more comprehensive coverage of meta systems by integrating with Delta Lake.
The message behind these features is clear: Confluent is no longer content with just providing the plumbing of the data pipeline — it wants a place at the AI table as intelligent systems become part of everyday infrastructure rather than isolated experiments.
Integrated into these new features is the Real-Time Context Engine, which manages structured data delivery to agents and arrives via MCP. The goal is to remove the need for brittle APIs and enable delayed batch updates to shift toward something more aligned with what agents will be expected to do. The Real-Time Context Engine is available in early access.
(Piotr Swat/Shutterstock)
The company also revealed Streaming Agents, a Flink-based environment that allows developers to create, test, and distribute agents directly on the platform. The initial agent implementations lacked the observability and debugging capabilities that Streaming Agents now provide.
Private Cloud offers the same capabilities behind the firewall, providing organizations that need tighter controls over data movement with built-in policy enforcement and improved replication. It also includes Tableflow’s support for Delta Lake, Unity Catalog, and Azure. These tools are aimed at making real-time pipelines to downstream analytics and AI tools much easier — without writing more ETL code.
However, the bigger question looms: how can AI agents act with intelligence if they’re always a few steps behind what’s actually happening? Even today, many systems rely on static snapshots, query layers appended to data lakes, or APIs that update too slowly to be of any benefit. When enterprises begin to automate their decisions — expanding these agents across business functions — that growing disconnect becomes a critical flaw.
Sean Falconer, Head of AI at Confluent, explains: “AI is just as excellent as context. The data is available to enterprises, but it’s frequently out-of-date, dispersed, or in a layout that AI can’t effectively utilize. Real-Time Context Engine achieves this by combining data processing, reprocessing, and serving live, converting persistent data flows into live contexts that enable faster and more constant AI decisions.” In a world of automated systems, context is not only useful — it is essential.
There’s a pattern that tends to define every wave of enterprise AI adoption. The innovation arrives first — then the reality check follows. Right now, that reality is setting in for agentic systems. The demand is there, but the basic architecture is still not prepared for what people hope these agents can do. While is easier to create an intelligent system, maintaining its trustworthiness, observability, and governance over time is a lot tougher.
(Deemerwha studio/Shutterstock)
That’s why the conversation is turning from algorithms to infrastructure. The companies that will define the next era of AI aren’t the ones coaching the most important models. They’re the ones figuring out how to keep the models connected to real data, without breaking when the business changes around them. It’s a quieter kind of progress — less prone to making headlines, but far more consequential.
“As AI-powered automated agents, assistants, and advisors begin to be used in organizations, curated, secured, compliant, and contextual data will be a key success factor in ensuring trusted outcomes,” states the IDC FutureScape: Worldwide Data and Analytics 2025 Predictions.
Only time will tell whether enterprise AI can evolve without rebuilding its foundations. Real-time context, governed pipelines, continuous feedback loops — these aren’t add-ons anymore, they’re prerequisites. Confluent’s push into this space reflects that recognition, and it’s one of the first serious signs that the industry is starting to take the “plumbing” as seriously as the intelligence sitting on top of it.
If AI agents are going to move from novelty to reliability, the future won’t be defined by how big the models get. It’ll come down to whether the systems feeding them are finally built for the pace of reality.
Related Items
The Quiet Rise of AI’s Real Enablers
Powering Data in the Age of AI: Part 3 – Inside the AI Data Center Rebuild
Unlock 5 Key Insights for Building High-Performance AI Infrastructure – From Power to Production
