Agentic Development for Business Central

A specification-driven approach to AI software engineering

Changing delivery with AI. Keeping it aligned with specifications.

Agentic AI is drastically changing how software is built. AI can generate code. AI assists with analysis, documentation, and speeds up engineering processes that developers face daily.

Development becomes faster with agentic AI. But faster does not automatically mean better.

To deliver reliable Business Central solutions that are easy to maintain over time, projects still need a clear understanding of requirements, well-defined solution architecture, effective validation, and knowledge that remains available long after go-live.

Adopting agentic AI makes alignment critically important. Specification-driven development (SDD) helps keep people, AI-assisted workflows, and processes aligned as they rely on the same approved specifications.

For Business Central projects, that means

◼️ clearer requirements and implementation guidance

◼️ less rework

◼️ better code quality and architectural consistency

◼️ more predictable delivery outcomes

The challenge: developing faster with AI does not mean doing it better

Many teams are actively exploring AI software engineering. They experiment with agentic coding and adopt AI tools for analysis, development, and testing. Yet most delivery problems do not happen because teams cannot write code.

They happen because information gets lost.

Requirements discussed in workshops evolve as they move through design, implementation, testing, and support. Different stakeholders may work from different documents, rely on different assumptions, or interpret the same business needs in a different way. Important decisions get lost in emails, meetings, and chats.

This is where discussions around agentic development and spec-driven development become really important.

The thing here is not whether AI can write code – it can. The thing is making sure humans and AI work from the same understanding of the business problem.

When there is no shared understanding, organizations meet familiar challenges:

unclear scope

changing expectations

inconsistent tech decisions and improvised design

duplicated effort and rework

outdated documentation

tribal knowledge that gets lost when key people leave

The more complex the Business Central solution, the greater the impact.

What is agentic development?

Agentic development relies on AI-assisted specialists (agents) to perform different tasks throughout the software delivery lifecycle: requirements analysis, solution design, implementation, review, testing, and documentation.

In practice, this may involve AI agents that operate within a set range of skills or defined boundaries.

With agentic development, speed is definitely a benefit, while misalignment is a risk. If contributors keep working from different assumptions, teams eventually get inconsistent results, but at greater speed.

At Global Mediator, we combine agentic software development with specification-driven development, with human oversight and control maintained at every stage of the delivery process.

A team of AI specialists under the hood

At Global Mediator, we don’t rely on a single AI assistant. We use multiple specialist roles that are engaged in all the stages of delivery. As a result, we offer structured agent orchestration: each specialist supports a specific part of the delivery process while staying connected to the same specifications.

◻️ Orchestrator: coordinates workflow stages and approvals, checks prerequisites, and logs the process to keep it transparent and to track artifacts easier.

◻️ Consultant: helps define business requirements and acceptance criteria.

◻️ Architect: designs the technical solution and implementation approach.

◻️ Developer: implements changes based on approved specifications.

◻️ Reviewer: performs independent validation and quality checks.

◻️ Documentation specialist: documents what was built – decisions, implementation details, and delivery outcomes.

Agentic AI relies on context

AI specialists make good decisions only when they rely on the same information as the project team. If such context is lost, even the best agents end up relying on assumptions and generic training data.

This makes MCP (Model Context Protocol) servers extremely important. They allow AI agents to get the right information and access the right tools at the right moment, meaning there is consistent access to shared project knowledge.

At Global Mediator, the knowledge layers for MCP servers cover:

Microsoft standards and best practices

Customer-specific requirements

Our internal engineering standards

Project guidelines and decisions

Individual working preferences

This way, AI agents can work with the project’s context, architecture decisions, and constraints.

What’s more, MCPs are platform-agnostic. So, the same approach works with GitHub Copilot, Claude Code, and other environments compatible with MCPs.

Making agentic development secure

Successful agentic development needs secure access to approved knowledge and workflows, not only powerful AI models. We use an internal VS Code-compatible extension to support this. Developers don’t have to configure AI tools independently. The extension provides a ready-to-use setup. It simplifies access to the approved MCP servers and allows us to securely distribute standardized AI agents across our projects.

Using the extension means our software developers use:

◾ approved engineering standards

◾ the latest project knowledge

◾ consistent workflows

◾ cohesive AI setups

With the extension, AI capabilities roll out in a controlled manner while maintaining consistency, traceability, and governance.

How specification-driven delivery works: orchestration in practice

Well-structured orchestration connects specialists, specifications, AI workflows, and human approvals throughout the delivery lifecycle.

◾ Define

Requirements and acceptance criteria transform into a shared specification. Human approval confirms the intended outcome before work moves forward.

◾ Design

The specification becomes a technical solution and implementation plan. Human specialists review architecture decisions before implementation begins.

◾ Implement

AI specialists build against approved specifications. Agent orchestration keeps implementation aligned with requirements, architecture, and project constraints.

◾ Review and Document Review findings, approvals, and documentation are linked back to the specification. The reviewer agent performs an independent validation of the produced code against the approved specifications.

Human oversight remains part of the final orchestration stage, keeping project traceability from requirement to delivery.

Benefits of adopting specification-driven development

Complete alignment between teams

Business and technical teams – stakeholders, consultants, architects, developers, and support – work from a shared understanding of the solution.

Less delivery risk and rework

Clear requirements and documented decisions help avoid misunderstandings before they become expensive project changes.

Stronger governance and traceability

Organizations can clearly see how requirements shape solution design and how design influences implementation.

Better future maintenance

It’s easier to enhance, upgrade, and support the solution since there is access to the original business and technical context.

More effective use of AI

Whether teams adopt spec-driven AI coding or broader AI engineering practices, specifications provide the context needed to keep delivery aligned with business intent.

Human in the loop

AI does accelerate delivery, but humans remain accountable. Reviews, approvals, and validation checkpoints help ensure that quality, compliance, and business objectives remain under control.

Build Business Central solutions with confidence

As agentic development becomes widely used, successful teams will need to keep people, specifications, and AI aligned around the same outcome.

Specification-driven development helps with that.

If you're planning a migration, modernization, or complex Business Central implementation, let's discuss how a specification-driven approach can help you combine the benefits of agentic AI with governance, traceability, and delivery discipline enterprise projects depend on.

You are very welcome to get in touch.

Please fill in the form to request details about the package / product / service you are interested in.