In the rapidly evolving Microsoft ecosystem, the conversation around AI has largely been dominated by flashy demos of Copilot and generative AI. However, for organizations looking to build maintainable, governed, and scalable business applications, the real challenge begins after the demo ends. The transition from a simple prompt to a production-grade enterprise experience requires a deep understanding of application architecture.
Sara Lagerquist, a Microsoft MVP and veteran applications architect with over 12 years of experience, suggests that the future of enterprise AI isn’t just about chatbots, it’s about how we structure the apps themselves. The debate between Model-Driven Apps and Canvas Apps has taken a new turn with the introduction of Generative Pages, a feature Lagerquist describes as a “chameleon” that bridges the gap between structured data and custom user experiences.
Model-Driven Apps: The Foundation for Enterprise AI
For many years, the choice between Model-Driven and Canvas apps was a trade-off between speed and customization. Model-Driven apps, rooted in the heritage of Dynamics 365, offer a “free” infrastructure. When you build a Model-Driven app, you aren’t just building a screen; you are getting enterprise-grade navigation, advanced search functionality, and a UI that is instantly recognizable to anyone familiar with Excel, Outlook, or the broader Microsoft 365 suite.
Lagerquist argues that for most business processes, Model-Driven apps should be the starting point. The overhead of a standalone Canvas app, configuring every button, layout, and data connection from scratch, can become “gruesome” as complexity grows. In contrast, Model-Driven apps allow for rapid development where the focus remains on the business logic and the data model rather than the pixel-perfect placement of UI elements.
The Rise of Generative Pages
While Model-Driven apps are efficient, they have historically been rigid. This is where Generative Pages (Gen Pages) change the game. A Gen Page allows developers to embed a fully custom, AI-generated UI within the structured environment of a Model-Driven app. This combination offers the best of both worlds: the robust backend of Dataverse and the flexible, dynamic frontend of a custom-built interface.
According to Lagerquist, Generative Pages solve several traditional pain points:
Reducing “Clickiness”: In a standard Model-Driven UI, creating many-to-many relationships or batch-creating records can involve five or six clicks per record. A Gen Page can condense this into a single, streamlined interaction.
Visual Flexibility: You can transform a standard list view into a Kanban board, a scheduling calendar, or a multi-table wizard without leaving the app environment.
Speed of Development: Building a complex custom UI with Power FX in a Canvas app takes significant time. With Gen Pages, AI handles the heavy lifting of the UI generation, allowing architects to focus on the outcome.
Maker Portal vs. Professional Coding Tools
One of the most critical insights for architects is how these pages are built. There are currently two primary paths for creating Generative Pages, and the choice between them significantly impacts the quality and maintainability of the final product.
The Maker Portal Experience
The Maker Portal offers a low-code entry point where users can prompt a page into existence using a chat box. While accessible, Lagerquist notes several limitations:
Model Constraints: It often defaults to older AI models (like GPT-4.1), which may not provide the most sophisticated results.
Inefficiency: Every time you refine a prompt to change a small detail, the AI tends to regenerate the entire page, which can be counterproductive.
Regional Availability: Many features are still in preview or limited to specific regions like the US and UK.
The Pro-Code Path: VS Code and GitHub Copilot
For enterprise-ready solutions, Lagerquist strongly recommends using Visual Studio Code (VS Code) in tandem with GitHub Copilot. Microsoft has released a specific “skill” or plugin for Gen Pages that allows for a much more controlled development process.
Using these tools allows you to choose your AI model (such as Claude or the latest GPT versions) and enter a Planning Stage. In this stage, the AI creates HTML wireframes that you can review and approve before any credits are spent on building the final code. This “solid way” of building is GA (Generally Available) worldwide and results in a more robust, professional-grade UI.
Strategic ROI: Changing the Architect’s “No” to a “Yes”
Historically, architects often had to say “no” to complex UI requests because the cost of development and long-term maintenance outweighed the business value. AI has shifted this Return on Investment (ROI) calculation. Because development is now “a million times quicker,” the barrier to entry for custom experiences has dropped.
“Before, I was an architect that pulled the brakes a bit more,” Lagerquist admits. “I would say ‘no’ a lot more before than I will now because of the speed of development of this feature.” This shift allows businesses to be more agile, adapting their apps to specific user workflows without the traditional technical debt associated with custom code.
Actionable Insights for AI App Architecture
To successfully implement AI-driven experiences in Power Apps, consider the following takeaways:
Start with the Data: Ensure your data is structured in Dataverse first. Generative Pages are most effective when they are surfacing and interacting with existing, well-governed data.
Identify UI Gaps: Use Gen Pages specifically where the standard Model-Driven UI fails to meet the user’s workflow needs, such as complex data entry or specialized visualizations.
Leverage the Planning Stage: Don’t just prompt and pray. Use tools like VS Code to iterate on wireframes before committing to a final build.
Think Beyond the Prompt: A good prompt is important, but a good architecture is vital. Use images and wireframes as context for the AI rather than relying solely on text descriptions.
Conclusion
The debate between Model-Driven and Canvas apps is no longer a binary choice. By leveraging Generative Pages, enterprise architects can build solutions that are both structured and highly customized. While the Maker Portal is excellent for proof-of-concepts, the true power of this technology is unlocked through professional development tools like VS Code and GitHub Copilot. As AI continues to accelerate development speeds, the role of the architect shifts from a gatekeeper of complexity to an enabler of high-value, tailored business experiences.