Agents of Data: Defining Agentic AI and Its Impact on Data Engineering
It is no surprise that agentic AI has emerged as one of the most talked-about aspects of artificial intelligence. But what does it really mean, and why does it matter for modern data teams?
In the first episode of the Agents of Data Podcast, Matillion’s Joe Herbert (Maia’s Principle Solution Architect) sat down with Julian Wiffen (Chief of AI & Data Science) and Sam Perrin (Senior Staff Software Engineer, AI and one of the architects behind Maia) to explore what agentic AI is, why context matters, and how it will transform the way businesses approach data engineering.
What is Agentic AI?
While definitions vary, agentic AI can be thought of as AI systems that don’t just process information, but also make decisions, orchestrate workflows, and coordinate with other agents or tools.
A single LLM process is like an individual contributor. An agentic system acts more like a manager, coordinating multiple contributors, choosing the right tools, and assigning specialized tasks.Julian Wiffen Chief of AI & Data Science| Matillion
This structure enables systems to handle both simple and complex work: from quickly scanning large volumes of text with lightweight models, to applying deeper reasoning with more advanced ones, all within a framework that adapts to the task at hand.
From Hidden Helpers to Impacting Enterprises
Many organizations are already benefiting from agentic AI without realizing it. Every time a user interacts with a large language model (LLM) that calls external tools, such as web search or summarization services, there is likely an agentic system behind the scenes deciding what to do next.
Coding is one of the clearest examples of this today.
Entire engineering teams are becoming ‘10x more efficient’ with coding agents like Claude Code or Cursor. In data engineering, Matillion’s Maia brings this same acceleration — enabling users to build and refine data pipelines through natural language and autonomous orchestration.Sam Perrin Senior Staff Software Engineer, AI| Matillion
Why Agentic AI Matters for Business Leaders
For Chief Data Officers and enterprise leaders, the real value lies in scaling limited resources. Every data team faces bottlenecks: more requests for data than capacity to deliver.
Agentic AI helps:
- Expand productivity: Junior staff can take on work with the support of AI ‘mentors’.
- Reduce cost: Organizations rely less on high-cost contractors or external GSIs.
- Accelerate time-to-insight: Data pipelines and integrations that once took months can now be delivered in days.
The most obvious opportunities are where bottlenecks are greatest: data engineering, compliance workflows, customer support analysis or any repetitive, high-volume process.
Trust, Risk and Quality Control
A natural question arises: if agentic systems are making decisions, how do we ensure quality?
The answer? It begins with a human-in-the-loop approach. Early deployments still require human approval, especially for high-stakes tasks like financial transactions. Over time, confidence grows through methods like:
- Sampling and review: Humans verify a percentage of AI-driven outputs.
- Agent-as-judge: One AI evaluates the work of another, ensuring consistency against defined criteria.
- Cross-agent validation: Multiple agents compare or “second opinion” on one another’s outputs.
Gradually layering the evaluation process allows businesses to scale while still maintaining oversight and accountability.
Agent-to-Agent Communication: A2A and MCP
The next wave of innovation is agents talking to agents. Standards like the Model Context Protocol (MCP) and Agent-to-Agent (A2A) specifications define how systems can discover, interact and share tasks with each other.
Think of MCP as the ‘USB port’ for AI systems. A standard way for agents to plug into new tools or services. A2A goes further, enabling agents to declare their roles, share status and execute tasks collaboratively.
For data teams, this opens the door to entirely new workflows. For example, a BI dashboard agent might request data it doesn’t yet have, triggering Maia to automatically build the pipeline, validate it and hand back the results, without a human ever writing SQL.
The Importance of Context
Effective agentic AI depends on context. Think of agentic AI as a new starter in your business. In order for them to perform well, they’ll need the right instructions, rules and business knowledge.
That’s where context files come in. Often written in Markdown, they provide structured background, schemas, naming conventions, and business definitions that agents can reliably interpret. By giving agents the same institutional knowledge that experienced team members use, businesses ensure consistency and accuracy in outputs.
A New Way of Working
As Julian and Sam highlight in the first episode of the Agents of Data podcast, success with agentic AI isn’t about waiting for perfect training courses; it’s about learning by doing. Experimenting with tools, refining context files, and iterating on workflows quickly builds capability.
For Matillion, this philosophy is embodied in Maia, the agentic data team inside the Data Productivity Cloud. Maia automates low-level engineering tasks, evaluates its own performance, and adapts to the unique data landscape of each customer environment.
The result? Data teams can focus less on firefighting pipelines and more on driving business impact.
Agentic AI represents a fundamental shift: from AI as a tool, to AI as a decision-maker within complex systems.
For data leaders, the opportunity is clear: reduce bottlenecks, cut costs, and make data engineering a strategic advantage.
Looking to catch the full episode?
Featured Resources
Agents of Data: Preparing Organizations for Agentic AI
Agentic AI has gone from curiosity to core strategy in what feels like a matter of months. But while the technology is racing ...
BlogAgents of Data: Digging into Semantic Layers
Semantic layers have quietly powered business intelligence tools for years. Now, as agentic AI systems emerge, they're ...
BlogHuman in the Loop in Data Engineering
Data pipelines are the backbone of modern analytics, but they're also notoriously fragile. The most resilient pipelines ...
Share: