Directions EMEA 2025: AI-first development for Dynamics 365 Business Central
Microsoft's message to Business Central partners was blunt: AI won't replace your team, but people who know how to use it will. The BC R&D group is expanding AI agent options for both functional consultants and developers, with a focus on shipping real tools-not hype.
"We are at the dawn of a new coding era," said Dmitry Chadayev. The latest models push past autocomplete. The new coding agent plans and executes work end to end from a natural-language goal.
From autocomplete to autonomous coding
- Creates AL source code and compiles it
- Runs tests on its own outputs
- Writes online documentation
- Commits to source control
- Understands approaches like MCP
In a demo, a consultant wrote detailed functional specs-behaviors, rules, outcomes, and tests. The agent generated tables, pages, extensions, codeunits, demo data, and even Copilot hooks inside BC in minutes.
What this means for consultants
You can move from vague ideas to working prototypes without waiting on a full dev cycle. Richer prototypes mean faster feedback, tighter specs, and fewer surprises when engineers step in.
What this means for developers
This is additive, not a replacement. The team set clear expectations for developer ownership:
- Guardrails and quality standards
- Hardening and optimization
- Solution architecture
- Reusable components
- Mentoring and code reviews
"This is great news for professional developers," said Steffen Balslev. With better intent and higher-quality specs from generative output, "My job remains the same [as a BC developer]." The focus stays on engineering excellence and durable solutions.
Tooling: AL-first, Codespaces, and more
The previewed tools are AL-specific, with more coming for testing, diagnosis, and bug fixing. Microsoft also showed a new setup: VS Code in GitHub Codespaces with the AL extension and access to MCPs.
Why native BC agents
- Context: Agents operate with first-class awareness of AL patterns, BC schema, and extension models.
- Tool access: Built-in hooks for compilers, tests, and source control keep the loop tight.
- Governance: Security, permissions, and auditability stay inside your existing BC and DevOps guardrails.
- Quality: Native scaffolding makes it easier to enforce standards and reuse proven components.
- Speed to value: Less glue code, fewer brittle integrations, faster iterations.
How to get value in the next 90 days
- Write agent-ready specs: behaviors, rules, acceptance tests, and edge cases in plain language.
- Stand up a Codespaces workspace with the AL extension and your BC repo.
- Create a patterns library (posting, dimensions, permissions, telemetry) for the agent to reuse.
- Define guardrails: coding standards, test coverage thresholds, security checks, and review gates.
- Pilot on low-risk customizations first; measure cycle time, defect rates, and review effort.
- Expand automated tests so agents can self-verify changes.
- Upskill consultants on writing clear specs and test cases that agents can act on.
Need a structured path to level up AI-assisted coding skills? Consider this practical track: AI Certification for Coding.
The takeaway
BC's AI agents are moving from helpers to doers. Consultants can ship credible prototypes faster, and developers can focus on architecture, quality, and reuse. Start building the workflows now so your team is ready as these tools roll out.
Your membership also unlocks:
