In 2026, data engineering isn't just about managing data—it's about building intelligent systems that power business strategy. Companies are moving beyond batch warehouses to real-time, cloud-native ecosystems that deliver insights instantly.
As AI and data engineering converge, pipelines now generate embeddings, feed vector databases, and power GenAI applications. Success depends on data quality, observability, and governance forming the foundation for trustworthy, scalable AI.
DBTA recently held a webinar, Top Trends in Data Engineering for 2026, with experts who discussed how this year marks a turning point: from pipeline building to engineering adaptive data platforms that drive speed, intelligence, and trust.
According to Sumeet Kumar Agrawal, VP product management, Informatica, “2026 is the year data becomes active.” He asked, “in a world of sub-second AI, is your data stack a bottleneck or an accelerator?"
It is a balancing act between performance and governance. A successful data engineering platform needs:
- Data Integration: Ability to integrate data from hundreds of sources and make it available in-time.
- Automated Governance: Moving security from a manual check to an embedded Data Contract.
- Predictive Scaling: AI-driven ops that adjust infrastructure before the bottleneck happens.
This is where agentic AI comes in, Agrawal explained. Agentic AI enables enterprises to achieve unprecedented levels of efficiency, accuracy, and scalability.
There are 3 pillars to building an enterprise agent:
Build accurate AI agents: Ground AI agents with trusted data. Leverage existing investments and unify the metadata system of intelligence.
Connect AI agents across the enterprise: Unified collaboration means recruiting all AI Agents—from anywhere. Streamline multi-agent workflows and scale impact across the organization.
Manage AI agents with confidence: Enable enterprise-grade security and control, along with full agent lifecycle management. Future-proof flexibility—supporting trusted framework for LLMs and standards.
Informatica AI Engineering enables users to build accurate agents, connect agents across enterprise, and manage agents with confidence, Agrawal said.
Right now, Oz Katz, co-founder and CTO at lakeFS said, data quality is a mess of excel and CSV files along with AI slop.
AI is changing quickly, Lucas Beeler, senior solutions engineer and architect at Aerospike, noted. Predictive AI predicts what will happen. Generative AI generates new artifacts. And agentic AI is all about planning and doing.
In predictive ML applications, such as fraud detection and high-frequency trading in capital markets, milliseconds can mean millions of dollars. In these applications, latency is king.
Prediction accuracy and generative fidelity all depend on access to ever larger amounts of data and the ability to deliver that data to the model quickly. Storage density and throughput rule the roost.
Data scientists do data analysis and build models that may not be production ready. Data engineers implement those models and their integrations at production speed and scale. MLOps deploys models, checks their health for accuracy degradation (“drift”) and redeploys updated models, maintaining the model lifecycle.
For the full webinar, featuring a more in-depth discussion, Q&A, and more, you can view an archived version of the webinar here.