Agentic AI is no longer just a passive assistant; it’s becoming an active collaborator. These AI agents are evolving into real team members, contributing meaningful value and accelerating development in ways that felt like science fiction just a few years ago.
TL;DR
Agentic software engineering marks a shift from using AI tools to actively collaborating with AI agents throughout the development process. These agents don’t just assist, they contribute code, optimize workflows, and drive productivity gains.
Tools like GitHub Copilot, Replit Ghostwriter, and Cognition’s Devin are leading this shift, enabling AI agents to create code, suggest architecture patterns, and even manage workflows alongside engineers.
According to Gartner research, by 2028, 75% of software engineers within enterprises will be using AI code assistants, a sharp increase from fewer than 10% in early 2023.
According to Business Insider, in 2025, 97.5% of organizations report using AI coding tools at work . More than 58% of companies report 20–50% productivity gains from these tools, with 25% seeing even larger improvements. While a similar study found that major tech companies such as Microsoft and Google are unsurprisingly early adopters and industry leaders in this new space, with AI-generated code now accounting for 20–30% of the new codebase, and at one firm, nearly 50% of new code is AI-generated .
Separately, a controlled study even found that developers using GitHub Copilot completed tasks 55.8% faster.
The takeaway is clear: AI agents don’t just assist. They collaborate, enhance, automate, and streamline. And one thing is certain: they’re here to stay.
But what does this mean for data teams?
After all, data engineers also write code. They collaborate in teams. They ship business-critical outputs. The difference? Instead of applications, they build data products.
This article explores how the principles of agentic software engineering are making their way into the world of data, and why the future of data engineering is agentic, too.
What is Agentic Software Engineering
Agentic software engineering describes a model where engineers work with AI, not just using it. These agentic tools:
Generate complete functions or modules
Catch and fix bugs
Recommend code optimizations
Manage workflows, documentation, and even test coverage
The goal? Faster time-to-value and better outcomes, not by replacing developers, but by amplifying them.
Agentic systems don’t just respond, they initiate. That's the leap from automation to collaboration.
Ian FunnellData Engineering Advocate Lead| Matillion
Agentic Software Engineering: A Definition
Agentic software engineering harnesses AI agents powered by large language models (LLMs) to automate and optimize every phase of the software development lifecycle (SDLC). These intelligent agents can interpret requirements, plan development tasks, generate and refactor code, run tests, and even manage deployment and monitoring, often without human intervention.
Unlike traditional, linear workflows, this approach introduces dynamic, adaptive, and decentralized processes. The result: faster delivery, fewer errors, and more time for engineers to focus on strategic, high-impact work. By embedding intelligence throughout the SDLC, agentic software engineering accelerates innovation while reducing the operational overhead of building and maintaining software.
Crucially, this shift isn’t hypothetical. AI-suggested code is being used more and more frequently as the models learn, while some tools now even feature autonomous issue resolution. This is just a glimpse of what the future of full AI/code collaboration will look like at scale.
Software Engineers Build Apps. Data Engineers Build Products.
There’s a strong parallel between modern software developers and modern data engineers. Both roles involve:
Teams working together
Writing and maintaining code
Shipping something that the business depends on
The difference lies in the output.
Software Engineering
Data Engineering
Applications that perform tasks
Data Products that inform decisions
End users that care about the experience
End users care about the insight
Code is invisible to the user
Pipelines are invisble to the user
In both cases, the end users don’t care about the backend. They only care about the output, and what it enables.
Agentic Data Engineering: AI Collaboration
If agentic software engineering is about building applications through AI collaboration, then agentic data engineering is about building data pipelines, and ultimately, data products, through AI collaboration.
Increased data analyst productivity by fulfilling data requests 60% faster.
4 x data requests fulfilled in 3 years
The exciting thing? This study was conducted before Matillion unveiled the AI agent Maia, the team of agentic data engineers…
AI in data engineering is a game changer.
Auto-generate SQL and Python for transformations
Suggest orchestration improvements
Detect schema drift or anomalies
Recommend pipeline enhancements based on lineage, usage, or past patterns
This isn’t just AI doing things; it’s AI that understands what you’re building and contributes.
The productivity gains and ROI promise to be something rarely seen before in the enterprise environment.
The Shared Goal: Data as a Product
Agentic collaboration makes even more sense when you consider the shift toward “data as a product”, a key principle in the Data Mesh model.
Just as software engineers deploy code to production, data engineers are shipping domain-owned, business-ready data. These data products must be:
Discoverable
Trustworthy
Well-documented
Actively maintained
In the end, the user doesn’t care if you use Matillion, dbt, raw Python, or any other model, method, or software. They care that the final data is accurate and actionable.
In software engineering, the end goal is an application that does something useful. In data engineering, it’s highly consumable data that helps the business operate.
Ian FunnellData Engineering Advocate Lead| Matillion
Matillion’s Role in Agentic Data Engineering
At Matillion, we see agentic AI not as a feature, but as an evolution in how data work gets done. Our platform is built to help data teams collaborate with AI to drive productivity, trust, and insight.
Meet Maia: Your Team of Agentic Data Engineers
Maia is your always-on workforce of agentic data engineers, built by the team that pioneered Cloud ELT. Collaborate with Maia to deliver data faster and automate the repetitive work, so your human engineers can put their expertise to better use.
Maia enables data engineers to:
Write SQL or Python steps with natural language
Interpret pipeline errors and debugging context
Accelerate development while maintaining governance and consistency
And much more
Instead of starting with a blank canvas, engineers start with a partner.
Built-in collaboration tools for team-based development
Lineage, observability, and orchestration baked in
Scalable, governed, and ready for enterprise needs
Oh, and Maia is available exclusively within the Data Productivity Cloud.
Agentic AI isn’t just about speed. It’s about enabling data teams to focus on what matters most: delivering value to the business.
But why now? The shift towards agentic systems is picking up pace, and at speed. It is increasingly being seen as a critical part of business infrastructure within the enterprise.
And those who do not keep up could well be left behind.
AI agents are only valuable if they improve trust, productivity, and time-to-insight. That’s what Maia has been designed to do.
Ian FunnellData Engineering Advocate Lead| Matillion
The Agentic Future of Data Work
Agentic software engineering shows us what’s possible when humans and AI collaborate, not just automate. The same principles apply to data teams.
Agentic Data Engineering is about building pipelines faster, yes, but more importantly, it’s about building better data products, accelerating insight delivery, and empowering engineers to solve more meaningful problems.
With tools like Maia and platforms like the Data Productivity Cloud, Matillion is helping shape this new era. An era where AI doesn’t replace the engineer, it amplifies them.
Agentic software engineering is a development model where AI agents work alongside engineers, not just as assistants, but as active collaborators, contributing code, optimizing processes, and managing parts of the software lifecycle.
Traditional automation follows static rules. Agentic AI can interpret context, adapt to changing requirements, suggest improvements, and even initiate tasks, much like a human teammate.
Both roles involve coding, collaboration, and delivering business value. Just as software engineers use agentic tools to build apps, data engineers can use AI agents to build and manage data products more efficiently.
Share: