On June 10, Mendix announced a repositioning—from a low-code platform to the Agentic Enterprise Platform. The question isn’t whether this is a marketing move. The question is what’s changing under the hood, and what that means for you in practical terms.
We’ve been working with Mendix as a strategic partner for years. We use it to build solutions for government, manufacturing, and mobility. And we’ve noticed that many articles about this shift get bogged down in hype. Terms like “autonomous systems” and “intelligent orchestration” are thrown around without anyone explaining what that actually means to an IT manager on a Wednesday morning.
This article is different. We’ll explain three shifts that we believe are truly relevant. We’ll do so in concrete terms, with examples. And we’ll wrap up with some actions you can take this month.
The core shift in one sentence
There is a fundamentally different design issue between low-code and agentic.
How can we build faster?
Business and IT come together on a single platform.
Visual modeling instead of coding line by line.
An application that would otherwise have taken half a year to develop is now live in six weeks.
Who monitors what happens independently?
AI agents make autonomous decisions, retrieve data from various systems, and initiate processes.
Sometimes without any human intervention.
The focus of design is shifting toward leadership and trust.
From Fixed Workflows to Adaptive Processes
In a low-code application, a process is defined in advance. Step 1 leads to step 2, with some conditional logic in between. This works well for predictable processes. It works less well for situations that are always just a little bit different.
An agent works differently. It’s given a goal—for example, “resolve this customer inquiry.” It then decides on its own which systems to consult, which data to combine, and what action to take next. That sounds magical, but under the hood, it’s a large language model that makes decisions within the parameters you’ve defined.
In practical terms, this means that processes that currently get bogged down in “exceptional cases” can now be handled. A citizen who asks a municipality a unique question. A logistics planner who has to make adjustments because a driver is out of commission. A service engineer who encounters a malfunction not covered in the standard manual. That’s where agents come in handy.
From Integration-Driven Development to Context-Driven Delivery
Many Mendix projects start with a question: How do we bring together data from ERP, CRM, and that one legacy database? In the past, you would have built a data integration layer for that. Mendix has even given it a name: the Data Hub.
With agentic, that integration layer doesn’t disappear. In fact, you still need it. Data has to come together somewhere, whether that’s in a data warehouse or a data integration layer. What changes is what happens on top of it. You now add an agentic layer on top that combines context, makes decisions, and executes actions.
Mendix solves this on its own with a Knowledge Graph: a single layer on top of existing systems that links the data without requiring you to move the underlying silos. ERP, CRM, document repositories, cloud storage, and big data remain where they are. The graph provides the context within which an agent can operate. This is an important architectural choice, because it means you don’t have to completely overhaul your entire data landscape before you can start thinking in terms of agents.
That’s a different kind of work. Less code, more policy formulation. Less “how do we extract data X from system Y,” more “in what situations can an agent autonomously make an accounts payable entry, and when does a human need to approve it?” Those kinds of questions.
Only 2 out of 10 IT leaders say their team is ready. That’s not alarming—it makes sense. No one builds something for the first time while already being ready for it.
From User Trust to User Control
This is what concerns us the most. And it’s something we don’t see enough of in the discussion.
In a traditional application, the user knows what’s happening. They fill out a form, click “Submit,” and see a confirmation. It’s linear and transparent. In an agent-based application, there’s an intermediate layer that the user doesn’t always see. The agent has consulted eight systems, made three decisions, and sent one email before the user sees any response.
This calls for a new design of human-machine interaction. The user must be able to see what the agent has done, understand why, and intervene if things threaten to go wrong—not through an “autonomy override” button hidden somewhere in a settings screen, but as an integral part of the user experience.
Mendix is building an Enterprise Trust Layer for this purpose, with four pillars that, in our view, are non-negotiable for agentic applications:
This sounds like compliance jargon, but in practice, it’s simply the prerequisite for feeling comfortable working with an agent.
We’re not new to this. Last year, we were already applying chatbots and RAG applications to specific challenges at organizations. While many companies were still exploring the possibilities, we were already implementing solutions. We’re now taking that same early approach with agents. We’re actively working on the first agent use cases—not because it’s a trend, but because we recognize the patterns.
That's not a feature. That's a requirement.
Five Phases for Agentic Adoption Without the Hype
From strategy to evaluation: an approach we’ve seen work time and time again. Each phase builds on the previous one. Skip one, and you’ll pay the price later.
Start by thinking, not by building
Define your vision and set specific goals.
Which processes are suitable for agentic?
Without a strategy, you're just building something that's destined for the trash.
One well-defined use case
Don't start with a platform-wide agentic strategy.
Start with one specific process where the current approach is stalling due to exceptions.
Compare the benefits of an agent with the governance costs associated with it.
Policy and Frameworks Prior to Rollout
Determine who is authorized to approve what.
Which systems agents are allowed to access and how to incorporate user control.
Compliance with the AI Act is strictly enforced. Agentic often falls into higher-risk categories.
Rollout to multiple processes
Only once you know how a single agent behaves under real-world conditions can you scale up.
The Mendix Competence Center plays a key role here: no longer just platform governance, but agent governance.
Learning and Adjusting
Agentic adoption is not a one-time project.
After each phase, evaluate: What worked, what didn't, and why?
Lessons learned form the basis for the next cycle. This way, each rollout is smarter than the last.
What specific steps can you take this month?
Not ten action items. Just three.
- Join the Mendix webinar on June 10. Not because you need to keep up with every product launch, but because this is a landmark shift for a platform you’re probably already using or considering.
- Identify one process where agentic would be relevant. No need for an ambitious strategy. Just one process that’s currently bogged down by exceptions, where an agent could add value.
- Define your stance on user control. How much autonomy will you give an agent, and how will you make that visible to the end user? You can have that discussion right now, regardless of technology.
Want to chat about what this means for you?
We regularly meet with IT leaders who want to understand this shift before basing strategic decisions on it.