Table of contents
Download the 2026 Gartner® Magic Quadrant™ for Data Integration Tools
Dark green abstract background with subtle gradient shapes and rounded corners.

The Agentic Enterprise Will Be Built by Data Engineering Teams

November 13, 2026
Blog
6 min read
Summarise with AI

Here’s what I think this Gartner® recognition says about where data engineering is heading.

Gartner has recognized Matillion as a Visionary in the 2026 Gartner® Magic Quadrant™ for Data Integration Tools. We believe the recognition reflects where the market is going: agents doing data engineering across the whole lifecycle, engineers governing it, and one data foundation, with the context to match, serving both analytics and the agents now running across the business. While we believe this placement speaks to the strength of Matillion’s vision for the future of data engineering, the real proof is how we’re executing on it alongside our customers every day. The report is available to download here.

Why This Matters, and What We’re Seeing in the Market

For most of its history, data engineering fed reports. Engineers built pipelines that delivered consistent, timely, accurate data primarily used to build two classes of data products: dashboards that track the performance of the business, and machine learning workloads that support predictive analytics and data-driven decision-making. That work still matters, and in fact is only growing in importance every day.

But something new is being added on top of it. AI development teams are building autonomous agents. Those agents have the logic, but they need data to take action. Providing the data products they need to complete their tasks falls to data engineers, who now work hand in hand with AI development teams.

That moves the role into the operational world. These data products now feed software that acts: serving customers, running finance and supply chain processes, making decisions inside the tools people use every day. Reporting and analytics are still theirs, and their responsibility has grown. A wrong number on a dashboard gets caught in a meeting; a wrong number handed to an agent gets acted on. And a customer’s personal data handed to the wrong agent is a critical breach of trust, not just a bug. Data engineering is becoming one of the most mission-critical functions in the enterprise, and the agentic enterprise will be built by these teams.

Ask a data engineer whether they delegate parts of their work to AI agents, and these days you’ll rarely hear no. They’ve had agents draft transformations, untangle stored procedures, or write unit tests. What data leaders want is bigger than that: whole pieces of work handed over, start to finish, on production systems. Very few teams are there yet. Data teams can’t meet the demand by hand, and they can’t meet it with agents they don’t trust. So what will it take for teams to trust agents with the reins?

The Foundational Capabilities of Agentic Data Engineering

Coding agents are already capable, but data engineering needs its own approach, because a pipeline isn’t finished when the agent stops working on it. It keeps running against data that shifts underneath it, often data containing PII about customers, patients or finances, governed by conventions that often live in a couple of people’s heads. For an agent to take on that kind of work, three things have to be true at once:

  • It has to be able to take on a meaningful degree of autonomy: 10 tickets worked in parallel by a team of agents that intelligently pull humans into the loop.
  • It has to know how your organization builds data, persistently across sessions.
  • It has to sequence production-grade components, and not generate probabilistically every time, so that every pipeline follows your policies and leaves an auditable record of where the data came from.

You can get surprisingly far with one or two of these, but not all the way. Over the next few weeks I’ll take each in turn.

How Much Can You Actually Hand Over?

Most AI help in data engineering still arrives one step at a time. Write this SQL, explain this job. The engineer remains the glue between steps: breaking the ticket down, testing, deploying, documenting. By our own analysis, writing code is about a fifth of the time between a request and a governed data product, so speeding it up just moves the bottleneck.

Autonomy works as a ladder. At the bottom, an agent handles one task inside a ticket. At the top, agents pick up work as it arrives and look after pipelines once they’re live, while people approve the plan and make the final call on production. That top rung is where we’ve aimed Maia. For Sophos, a Maia customer, a pipeline that took 10 to 16 hours end to end now takes 2 to 5.

Context That Builds Up Instead of Resetting

Open a fresh session with a general-purpose coding agent and ask it to add a new source to your warehouse. Before it can do anything useful, you’ll be explaining your naming conventions, which system is the source of truth for customers, and how your team handles data that arrives late. Tomorrow you’ll explain it all again.

A schema tells an agent what columns a table has, not that two business units count customers differently or that every new table needs an owner before it ships. Without that knowledge, an agent makes the same reasonable-looking mistake a hundred times, and someone has to catch it a hundred times. A catalog records what your data is; agents also need a record of how your organization builds it. In Maia, that’s the job of the Context Engine. At PMG, a clinical research organization built from more than 20 acquisitions, understanding an unfamiliar pipeline now takes about half an hour instead of two days.

Built from Parts That Already Work

Agents that write every pipeline from scratch often produce perfectly decent code. Consistency is where they struggle. Ask twice and you get two slightly different pipelines, each of which someone has to read before it can be trusted. When an agent assembles pipelines from components already proven in production, the same request produces the same pipeline, and review can focus on what’s actually new, the business logic. This standardized implementation of enterprise-grade ETL components has been a key enabler of autonomy for Maia customers; for example, more than 95% of Sophos’s agent-built pipelines now deploy successfully for the first time.

Consistency is also what makes governance possible. When a team builds pipelines with a general-purpose coding tool, nothing in the generated code tells an auditor where the underlying data came from or what touched it along the way. When agents build from governed components, the platform knows what every step does. It can apply your data-handling and access policies while the pipeline is being built, and record lineage as a by-product, so you can show exactly where a number came from.

What This Means for Data Teams

Once the three are working together, the role expands further. Agents carry more of the building, so data engineers spend more time on the work their new responsibility demands: deciding how pipelines ought to be built, working with AI development teams on the data products agents need, and checking that the results can be trusted. Balfour Beatty described it better than I could: “Right now, 80% of my team’s time is spent on routine BAU demand. Maia gives us the freedom to flip that - so I get 80% of my team doing the transformative stuff that actually makes a real difference to our business.” This is particularly impactful for executing against enterprise AI initiatives, which rely on data that is well governed and well described, delivered at a rate faster than people can prepare by hand.

I believe the teams that win the next few years will be the ones that can hand whole pieces of work to agents and trust what comes back. That takes autonomy, context and tested components working together, and it’s what we built Maia to do. We believe the Gartner placement tells us we’re building in the right direction. Mission Control and the Context Engine only became generally available in July, so the proof comes from customers in production.

There’s a great deal more to build, and my focus from here is getting Maia into the hands of every data team whose backlog is outgrowing its headcount. Judge us on how quickly our customers get governed, trusted results out of Maia, because that’s the standard I’ll be holding us to.

My thanks go to the customers who put Maia into production early. If you’d like to find out whether any of this holds up on your own work, bring a real pipeline and your standards to a Maia workshop and we’ll measure it together.

Disclaimer and Sources

Gartner, Magic Quadrant for Data Integration Tool, Thornton Craig, Robert Thanaraj, Sharat Menon, Jenna Goodrich, Nina Showell, 2 November 2026. Gartner does not endorse any company, vendor, product or service depicted in its publications, and does not advise technology users to select only those vendors with the highest ratings or other designation. Gartner publications consist of the opinions of Gartner’s business and technology insights organization and should not be construed as statements of fact. Gartner disclaims all warranties, expressed or implied, with respect to this publication, including any warranties of merchantability or fitness for a particular purpose. Gartner and Magic Quadrant are trademarks of Gartner, Inc., and/or its affiliates. The Gartner document is available upon request from MATILLION - CO.

Read the 2026 Gartner® Magic Quadrant™ for Data Integration Tools

See why Gartner recognized Matillion as a Visionary, and where data integration is heading.
Soft yellow abstract background with smooth gradients and rounded edges.
Last updated
November 16, 2026
Sanjay Krishnamurthi
Chief Development Officer, Matillion
Sanjay Krishnamurthi is Chief Development Officer at Matillion, where he leads the product and engineering organizations behind Maia, Matillion's AI Data Automation platform. He has spent his career building the technologies behind modern enterprise data. As Chief Technology Officer at Informatica, he helped shape the modern data integration market. He later joined Microsoft, where he led the transformation of Azure Data Factory and Microsoft Purview into leading cloud data services. Most recently, as Vice President of Engineering at Google, he led the company's Data Integration and Governance services and helped pioneer their evolution toward AI-native architectures.
View Profile

Data management
made effortless

Enjoy the freedom to do more with Maia on your side.
Abstract dark teal geometric shapes background with diagonal lines and subtle gradients.