Episode 519: The Last Frontier: Can AI Finally Conquer BC Report Layouts?
Dynamics Corner episode 519 interviews a Business Central MVP developer about AI-assisted BC development with GitHub Copilot and Claude. Topics: the expense management agent (about 90% AI-written, about 6 months from idea to release), the BC quality agentic code review workflow, multi-agent workflows, AI for report layouts (good-looking Word layouts, but field mapping not yet solved; layout and conversion attempts rated 60-70% successful), and telemetry analysis via MCP.
- Introduction and guest background
- Neil Johnson's background and role as MVP
- Expense management agent and AI-developed features
- BC quality workflow and agenting in development
- AI development timeline and partner adoption challenges
- Partner investment and workflow consistency
- Personal AI tool experience and GitHub Copilot
- Game-based AI learning and code generation workflow
- GitHub Copilot setup and usage statistics
- Agent-based development workflow and multi-model exploration
- Agent review process, documentation automation, and Playwright
- Time management and productivity gains with AI tools
- German market work and AI performance with report layouts
- Check formats, customer requirements, and report layout evolution
- Claude AI capabilities for Word layouts and reporting
- RDLC to Word migration strategy and AI performance analysis
- Telemetry tools, MCP integration, and AI agents for analysis
- Telemetry benefits for partners and development teams
- AI tools beyond Business Central: Claude, GitHub Copilot, and personal applications
- Cognitive effects of AI reliance and future of programming languages
- Closing remarks and presenter contact information
Overview
The guests discuss how AI is changing Business Central development. They cite the expense management agent, described as about 90 percent written by AI with roughly 6 months from idea to release. They also describe the BC quality workflow, and one developer’s own use of GitHub Copilot for AL, with about 95 percent of code generated. The workflow uses planning, working and reviewing agents, and agent-generated documentation with Playwright screenshots.
The second half focuses on report layouts. The speaker says AI handles RDL reasonably well because RDL is HTML-like, with a 60-70 percent success rate in his tests. AI can produce good-looking Word layouts, but he has not yet succeeded in mapping fields. The discussion also covers Microsoft moving standard documents from RDLC to Word layouts, with RDLC still supported. It ends with telemetry analysis through MCP and the effect of AI on programming habits.
Key points
- The expense management agent was described as about 90 percent written by AI, with about 6 months from idea to release.
- Microsoft’s BC quality workflow adds agentic triage and code review with community instructions, and partners can add their own instructions.
- One developer reports about 95 percent of his code is written by GitHub Copilot and says it is less buggy than his manual code. He uses MVP access to the Pro Plus version.
- Multi-agent workflow: a planning agent, a working agent and a reviewing agent, with the output always reviewed manually. Switching between models burns through credits quickly.
- Advice for starting with Copilot: begin without instruction files, sub-agents or skills, and add them after learning the basics.
- Asked about RDLC-to-Word conversion, the guest rated his AI results about 60-70 percent successful. Generated RDL sometimes does not compile and Word field mapping has not worked yet, but the output is a good starting point.
- Claude can design an appealing Word invoice layout from an existing BC invoice. The missing link is connecting the fields properly.
Features
Every feature below: status not stated.
| Feature | At |
|---|---|
| Expense management agent | 6:34 |
| BC quality workflow with agenting and code review | 8:08 |
| GitHub Copilot for AL development | 14:40 |
| Multi-agent workflow for development | 19:19 |
| Instruction files and sub-agents for Copilot | 18:52 |
| Automated screenshot generation with Playwright | 24:14 |
| Agent-driven documentation generation | 22:00 |
| AI assistance for RDL report creation | 29:33 |
| AI assistance for Word layout design | 29:49 |
| Microsoft shift from RDLC to Word layouts | 28:49 |
| RDLC layout continued support | 36:40 |
| AI-powered performance analysis for profiler data | 37:25 |
| XRechnung and ZUGFeRD e-invoice support | 28:36 |
| Telemetry analysis through MCP | 40:26 |
| AI-powered proactive telemetry monitoring | 41:37 |
| Telemetry for code refactoring identification | 42:33 |
| AI development assistants (Claude, GitHub Copilot, Gemini) | 44:05 |
AL objects mentioned
As heard in the captions. A name that matches one object page by exact type and name links to it; the others stay as named.
- other “BC quality” at 8:23
- other “Playwright” at 24:25
- other “RDL” at 29:33
- other “Word layout” at 29:00
- other “RDLC” at 32:28
- other “XRechnung” at 28:36
- other “ZUGFeRD” at 28:36
- other “KQL” at 40:26
- other “MCP” at 40:26
Quotes
- 6:34 “the expense management agent. And I’m looking forward to seeing it and to seeing how it actually works.”
- 6:59 “it was developed completely using AI and it was developed at faster pace than the other features”
- 9:42 “the community instructions, and then partners can put their own instructions in there as well to help monitor the code”
- 10:04 “the expense agent was written I think it was 90% by AI was what they had mentioned”
- 10:20 “the period of time it took to develop was 6 months. And that period of time or from like idea to release, I think.”
Disclaimers in the video
- 6:59 coming-later: I’m really wondering how will it be accepted and how will it work out at the end you know in the wild when it’s released in the wild
- 29:23 preview: that area in which AI help to create new layouts is not yet fully covered in AI workflow yet. But I think that might also change
- 30:04 preview: the workflow is not 100% covered yet because I think the AL development is already covered. And this area is not yet covered.
- 33:29 preview: The mapping in word layout I haven’t succeeded yet. we’re not made the complete journey yet.
- 40:40 coming-later: set that up next week um is is my plan. Uh cuz I’m going to bring back an initiative
Presenters (as heard): Neil Johnson, Chris, Brad, Kristoff, Tynan Stark, Jeremy, Twilio, Torben, Milan, KB, Tine.
