Agents of Data: Becoming AI Native

AI-native companies are not just using AI; they are rebuilding their entire operating model around it. In this episode of the Agents of Data Podcast, Matillion’s Julian Wiffen (Chief of AI & Data Science) and Sam Perrin (Senior Staff Software Engineer, AI) explore what it takes to make the leap from experimentation to becoming truly AI-native.

The difference between a company that uses AI and one that is AI-native is becoming starker by the day. While many organizations are still stuck in the "chatbot phase", deploying isolated pilots that don't talk to each other, a new breed of AI-native companies is emerging. These organizations are winning on speed, scale, and resilience by embedding agentic AI into the core of their engineering, data, and analytics workflows.

In Episode Six of the Agents of Data podcast, Julian and Sam break down this transition, explaining why Chief Data Officers (CDOs) must lead the charge and how the role of the data team is fundamentally changing.

Defining the AI-Native Organization

Being AI-native isn't about having a flashy generative AI feature in your product; it’s about how your organization operates internally. An AI-native company uses agents to augment human capability across every function, from operations to engineering.

The conversation highlights a critical shift:

• Traditional companies treat AI as a tool for specific tasks (e.g., "write this email" or "summarize this PDF").
• AI-native companies build agent-driven workflows. They deploy coding agents to handle routine development, automated researchers to gather market intelligence, and context-aware copilots that understand the specific nuances of the business.

As Julian and Sam note, the goal is to move from "early experimentation" to systems that materially expand output. It’s the difference between a developer using an LLM to write a function and a developer managing a team of AI agents that implement, test, and document entire features.

Knowledge Curation is the New Coding

One of the most profound insights from this episode is the shifting value of work. In an AI-native world, knowledge curation becomes as important as writing code.

AI agents are only as good as the context they are given. If an agent doesn't understand your business rules, your data definitions, or your architectural standards, it will produce generic (and often useless) results. This elevates the role of the "knowledge curator", the person responsible for maintaining the context files, semantic layers, and documentation that feed the agents.

Instead of spending hours writing boilerplate code, engineers and data professionals in AI-native organizations spend their time:

• Defining the "Box": Creating the constraints and guardrails within which agents operate.
• Curating Context: Ensuring the agents have access to up-to-date business logic and data schemas.
• Reviewing Outcomes: Acting as the "human in the loop" to validate agent decisions rather than making every decision themselves.

The "Evaluation Gym": How to Trust Your Agents

How do you know if your AI is actually getting smarter? In traditional software engineering, a unit test either passes or fails. In the probabilistic world of AI, the answer is rarely black and white.

Julian and Sam discuss the necessity of building an "Evaluation Gym", a rigorous testing environment designed to benchmark agent performance. Because LLMs are non-deterministic (meaning they might give slightly different answers each time), you cannot rely on simple "vibes-based" testing.

The Evaluation Gym allows teams to:

• Run Automated Scenarios: continuously test agents against hundreds of real-world data scenarios, ranging from easy to complex.
• Measure Progression: Track metrics over time to ensure that fixing a bug in one area doesn't degrade performance in another.
• Build Confidence: Move from "hoping" the agent works to statistically proving it is ready for production.

Without this gym, organizations are flying blind, unable to distinguish between a lucky guess and genuine intelligence.

The Strategic Imperative for CDOs

The shift to AI-native operations is not a bottom-up change; it requires top-down leadership. The podcast emphasizes that CDOs and data leaders are the natural owners of this transformation.

Why? Because AI needs data. The organizations that will succeed are those that treat their data infrastructure as the fuel for their agentic workforce. Leaders must look beyond the hype and focus on:

• Targeting high-value use cases: Don't sprinkle AI on everything. Find the bottlenecks where agentic workflows can deliver 10x productivity gains.
• Culture of reinvention: AI-native companies embrace the idea that the way we worked yesterday might be obsolete today.

Ready to become AI-Native?

There is a palpable urgency in the discussion. The gap between AI-native companies and traditional enterprises is widening. To hear the full discussion on how to transform your organization's approach to AI, check out the latest episode.

Book a Maia session to experience the agentic data team.

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?

• Watch Agents of Data Episode 6 on YouTube here.
• Listen to Agents of Data Episode 6 on Amazon Music here.
• Listen to Agents of Data Episode 6 on Apple Podcasts here.
• Listen to Agents of Data Episode 6 on Deezer here.