Dynamics Business Central / NAV Developer Digest - Vol. 532

ArcherPoint’s Developer Digest focuses on Microsoft Dynamics 365 Business Central and Dynamics NAV development. This week’s volume includes context in AI prompts using Microsoft’s MCP, using AI to improve code quality, controlling memory usage in Azure Logic Apps, and limitations with BC Cloud.

The Dynamics 365 Business Central community, consisting of developers, project managers, and consultants, collaborates across various platforms to share valuable insights. At ArcherPoint by Cherry Bekaert, we greatly value their dedication and expertise. To ensure widespread access to this technical knowledge, we created Developer Digest.

Providing context in AI prompts using Microsoft’s Model Context Protocol (MCP)

Large Language Models (LLMs) like GPT or Claude are trained using massive amounts of data—but their knowledge is general and based on patterns. That means that prompts without specific context will return results based on general patterns that “look right”. In other words, they sometimes hallucinate.

That’s why context is important when creating prompts for AI. These prompts include additional information, such as rulebooks, coding style guides, limits on character/word length, and real-world working examples, to keep the AI response “grounded.” (See our blog on prompt engineering for Copilot.)

To help LLM systems produce more meaningful responses, AI industry leaders are adopting the Model Context Protocol (MCP) architecture. This open protocol provides context to LLMs by connecting AI applications to specific data sources and tools.

Microsoft recently announced its version of MCP in Copilot Studio, and GitHub offers repositories for many of Microsoft’s MCP servers.

Dmitry Katson offers an example of prompting the MCP for Microsoft’s Learn Docs with and without context in his blog. Without context, the response came back with several hallucinations. With context, the response leveraged the Microsoft_docs_search tool across Microsoft Learn, Azure Docs, M365, and more to generate an accurate response.

Check out Dmitry’s blog, Microsoft Learn MCP is LIVE!

Using AI to improve code quality and structure

Speaking of context in AI, several ArcherPoint consultants and developers met recently and discussed the practicality of using AI to generate AL code. For demonstration purposes, they used Claude from Anthropic.

They found that Claude was able to:

• Generate a loyalty program in AL from a brief prompt
• Create folders, objects, and helper functions (e.g., customer card extensions, enums, and loyalty point actions)
• Ground itself using a master prompt file to improve naming conventions and structure
• Write a product requirements document (PRD) from a high-level plan
• Commit code to Git

While Claude improved with context documents (grounding) and generated well-structured files and folders, it also generated different code each time and missed some object dependencies.

Ultimately, they concluded that Claude and similar AI tools are promising for reducing repetitive tasks and speeding up development, but still require strong human oversight and detailed input.

Controlling memory usage in Azure Logic Apps

It is common to process blob files (JSON, XML, CSV, etc.) when integrating with applications like Business Central, and Azure Logic Apps is a common way to handle them. The standard Azure Logic apps plan offers a fixed fee for the hosting plan accommodating multiple Logic App instances.

However, because the built-in connectors load all the information into memory by design, large blob files can fill up your available memory on the hosting platform and generate an Out of Memory exception.

Stefano Demiliani discusses how to monitor your memory usage and offers several options to mitigate the problem if available memory becomes exhausted.      

Read his blog, Azure Logic Apps: are you handling large blobs? Keep memory usage under control, to find out more.

Know your limitations with BC Cloud

Business Central Cloud provides a platform that offers powerful capabilities and a seamless ERP experience. However, there are limitations. IT administrators and developers should be aware of how the Business Central Cloud infrastructure is defined and monitor it regularly to ensure they do not exceed its limitations.

Saurav Dhyani discusses the Business Central Cloud components, such as the Tenant ID, various environments, and quotas, along with how to extend the platform’s capabilities in his blog, Understanding Business Central Cloud Infrastructure: Tenant ID, Environments & More!

Are you interested in Dynamics NAV and Business Central development? Check out our collection of NAV/BC Development Blogs.

Read “How To” blogs from ArcherPoint by Cherry Bekaert for practical advice on using Microsoft Dynamics NAV and Dynamics 365 Business Central.

Trending Posts