AI and Low-Code: Are We Ready for a Visual Co-Creation Lab for Applications?

How software development is becoming an increasingly collaborative and visual experience

Discover how the integration of Artificial Intelligence and low-code is transforming the way applications are designed and developed, making the process more collaborative, faster, and accessible to all stakeholders in the software lifecycle.

Today, Artificial Intelligence presents itself as an ally—a design companion capable of suggesting solutions and automatically generating models and logic. While many are attempting to use it in traditional development, it becomes challenging in that context to detect logical errors or flaws often hidden within thousands of lines of code.

This is exactly where low-code technologies like Mendix put the human at the center: within a visual, transparent, and easily modifiable environment, designers can build applications while leveraging AI features to examine, understand, refine, and continuously improve domain models, application flows, and functionalities. These naturally emerge and evolve through visual modeling. In this way, low-code proves to be the technology best suited to enabling true symbiosis between humans and AI—facilitating a co-creation process that combines automation, control, and human understanding.

Low-code provides pre-built, tested components that help build high-quality software while reducing errors. The result is a system that is immediately visible, inspectable, and easy to understand and validate.

In this context, having AI integrated into the IDE allows those designing and building applications to interact—throughout the Software Development Life Cycle (SDLC)—with advanced Large Language Models (LLMs), trained on a knowledge base managed by the technology provider, and potentially expandable to include the customer’s local LLMs. This knowledge includes domain models, entities, attributes, associations, and best development practices.

Everything happens naturally through a visual and conversational interface that allows the AI to be queried in parallel with creation. Users can ask for suggestions or clarifications, make changes, optimize, or automate when needed.

 

A Real-World Example: MAIA 

At WEGG, we are analysts who support clients in their digital transformation journey, translating their needs into application flows and requirements that are then implemented using low-code technologies—with all the benefits of speed, standardization, and development automation.

Whether it’s single applications or multiple solutions for diverse needs—managed by us or in collaboration with the client’s internal development team—we have seen how AI-assisted development within low-code platforms significantly improves development time and quality.

Specifically, we work with Mendix, a low-code technology recognized as a leader by Gartner, which has integrated an AI assistant called MAIA (Mendix AI Assistance) with three core support goals:

  • Guide: Helps developers and analysts by enabling them to ask questions and receive immediate, detailed responses via chatbots that provide summaries, flow insights, and suggestions on how to best build applications.
  • Assist: Offers real-time recommendations, ensures adherence to best practices, and automates repetitive tasks that slow down traditional development, like bulk updates.
  • Generate: Leverages generative AI to accelerate development by creating both front-end and back-end logic, generating application components, and automating tasks such as documentation writing, SQL command generation, and test data creation.

 

Supporting the Entire Development Lifecycle 

As mentioned, AI-assisted development in Mendix is seamlessly embedded within a unified, visual, and conversational interface. It spans the entire software lifecycle—from ideation to implementation—including deployment, testing, governance, monitoring, and feedback. Here are some examples.

Ideation Phase 

In the ideation phase, AI can assist in creating a prototype by generating a first version of the application from a requirements document (in PDF format). This document is uploaded into the conversational interface, where specific, contextual prompts can also be added to refine and guide the result more precisely.

With the Domain Model Generator, you can describe to MAIA the data you want the domain model to capture. The description can be detailed (e.g., "Create an entity ‘A’ with attributes ‘B’ and ‘C’") or more generic using contextual prompts.

This generates a foundational structure that can be refined over time. You can add attributes to existing entities, ask MAIA to explain or improve the domain model, and automatically link it to the back-end—without manual drag-and-drop actions. This approach is particularly useful for analysts and designers as it saves significant time during domain modeling while supporting the identification of the correct entities for the application.

MAIA remains accessible at all times via a dedicated chat interface—whether for generation, refinement, or other tasks. It can retrieve supplemental information or suggestions from Mendix documentation, forums, and the academy’s learning paths, organizing it into user-friendly formats (markdown, bullet points, etc.) with internal source references.

Designers can also use the AI to organize page content by inserting data views using widgets such as text fields for entity attributes. By uploading an image (screenshot, wireframe, or sketch) and adding textual instructions, users can request a layout replication to instantly preview the information presentation.

 

Application Structuring Phase 

During the application structuring phase, Mendix allows management of user stories through Epics—collections of related user stories representing larger work blocks, which are broken down into smaller, manageable tasks using a Kanban model. This helps organize and track progress of broader functionalities or initiatives.

Spesso chi scrive/condivide le user story ha poca esperienza nella stesura di testi tecnici; per questo, utilizzando la funzione MAIA Create User Story, è possibile generare un testo, aggiungere criteri di accettazione, modificare e perfezionare le storie. Questo approccio favorisce la collaborazione tra figure meno tecniche, permettendo loro di partecipare attivamente con un linguaggio naturale. 

User stories—complete with acceptance criteria and test instructions—can then be implemented directly into the domain model with AI support. These suggestions must always be validated and visually reviewed but help accelerate feature integration into the application model.

Additionally, integration between the IDE and the developer portal allows stories to be marked as ready for testing or completed, automatically updating the Kanban board.

After testing, version control can be applied across all entities for bulk updates, and the AI can generate documentation (even in multiple languages). These outputs require review but substantially boost productivity.

 

UX Improvement and Flow Understanding 

Often, application areas are more functional than aesthetically refined. Developers typically start from Figma wireframes, widely used by UX designers. These designs can now be used to automatically generate application pages, avoiding manual replication of components. The system leverages the application’s underlying structure to generate the necessary components, which can then be reviewed and customized.

Additionally, when joining an ongoing project, developers can explore microflows to understand application logic and ask MAIA for clarifications—e.g., how a purchase order flow is structured. It’s even possible to generate flows from PDF forms, which the AI transforms into HTML pages to serve as a starting point—especially useful for legacy format migrations.

If the PDF lacks buttons or other interactive elements, AI can create interaction schemes based on HTML examples. MAIA generates everything—CSS, JavaScript, etc.—to efficiently build a complete front-end.

Translation is also part of UX: with AI support, system texts can be automatically translated based on the app users’ language. While these translations should always be reviewed for localization and context, much of the work is already done.

 

Testing and Deployment 

AI also proves highly effective during testing and deployment, phases, where it helps automate repetitive tasks such as version control and integrating best development practices during code review.

For example, when a detailed commit message is provided, AI can apply and track bulk updates throughout the project’s versioning. A CI/CD pipeline in Mendix allows AI to perform reviews, offer best practice-based recommendations, detect anti-patterns, and—if critical thresholds are exceeded—block deployment with instant feedback provided directly in the pipeline.

This ensures that applications are only deployed when all quality conditions are satisfied, guaranteeing a secure and well-governed release. In this way, AI helps development teams maintain high standards across the entire lifecycle. implementare e mantenere nel tempo le migliori pratiche di sviluppo, garantendo che l’applicazione cresca sempre mantenendo un alto livello di qualità. 

Visual and conversational development, made possible by platforms like Mendix, is undoubtedly the future of software development. Analysts, designers, and developers can now significantly speed up their work—not just during ideation and design but also in structuring, testing, and deploying applications. This ultimately frees up more time to focus on refining requirements and improving the application itself.

The intelligent integration of AI like MAIA is transforming how we build software, enabling real symbiosis between human and artificial intelligence. In this evolving landscape, AI agents are no longer just support tools but true co-designers that automate repetitive tasks and suggest optimal solutions.

Looking ahead, we are moving toward a model in which feedback can be directly integrated from the user story level—through the approval and refinement of proposed flows and models.

02-s pattern02

Would you like to improve service management?

CONTACT US FOR A
CONSULTATION!