Introduction – Intelligence Alone Isn’t Enough
Artificial Intelligence is changing the way we think about software development, consulting, and business applications. Every week brings a new Large Language Model, a new AI framework, or another breakthrough in agentic AI. The pace of innovation is incredible, and it’s becoming increasingly clear that AI will play a significant role in the future of Microsoft Dynamics 365 Business Central. Like many professionals in our community, I was curious about how AI could improve the way I work. My journey started with simple experiments—testing different AI models, comparing responses, refining prompts, and exploring how AI could assist with everyday Business Central activities. Initially, I focused on understanding what these models were capable of and where they could genuinely add value to consultants and developers.
As my experiments evolved, I began building AI solutions specifically for Business Central. What started as a Documentation Agent soon expanded into UAT Agents, Troubleshooting Assistants, Code Review Agents, AI Knowledge Hubs, and eventually the broader vision of an AI First Business Central Practice. Each new project pushed me to think differently and revealed new possibilities for how AI could support every stage of a Business Central implementation.
At first, I measured success purely by capability.
Could the AI generate high-quality documentation?
Could it review AL code and identify potential improvements?
Could it create meaningful UAT scenarios?
Could it explain Business Central functionality in a way that consultants and users could understand?
Most of the time, the answer was yes. The results were often impressive, and every new experiment reinforced the feeling that AI had enormous potential. However, once I started applying these agents to real Business Central scenarios, I realized that capability alone wasn’t enough.
An AI assistant could produce technically correct answers, but that didn’t automatically make it useful in a customer implementation. Business Central consulting is rarely about producing the fastest answer. It is about understanding business processes, evaluating different possibilities, asking the right questions, and making decisions based on experience and context.
That was the moment my perspective changed. Until then, I had been focused on building more intelligent AI assistants. But I soon realized that intelligence alone wasn’t enough. An AI assistant could generate accurate answers and still fail to become useful in a real Business Central project.
The real challenge wasn’t building an intelligent AI assistant.
The real challenge was building an AI assistant that Business Central consultants could trust.
In consulting, trust matters more than intelligence. Consultants don’t just need answers—they need confidence that the recommendations are relevant, practical, and aligned with the customer’s implementation. They need an assistant that knows when to ask questions, when to investigate further, and when not to make assumptions.
That realization completely changed the way I design AI agents.
Instead of asking, “How intelligent is the AI model?” I started asking, “What would make an experienced Business Central consultant trust this AI assistant enough to use it during a real project?”
That single question became the foundation of every AI agent I’ve built since then.
Designing Specialized AI Agents
When AI first became mainstream, prompt engineering became one of the hottest topics. Everyone was searching for the perfect prompt, believing it was the key to building better AI assistants.
My experience led me to a different conclusion. Prompts are important, but they are only one small part of building a trustworthy AI agent. Long before writing the first prompt, I now focus on designing the agent itself. I ask questions such as: What business problem is this agent solving? Who will use it? What Business Central knowledge does it need? Which tools should it have access to? How should it investigate a problem? What should a successful outcome look like?
Another lesson quickly became clear. One AI assistant cannot solve every Business Central challenge. Initially, I tried building a single assistant that could handle documentation, code reviews, troubleshooting, testing, development, and support. While it could do a little of everything, it wasn’t exceptional at any one task.
That’s when I changed my approach.
Instead of one general-purpose assistant, I started building specialized AI agents, each with a clear responsibility, dedicated instructions, structured workflows, and access to the right Business Central knowledge and tools. A Documentation Agent thinks differently from a Troubleshooting Agent, and a Code Review Agent requires different guidance than a UAT Agent. The results improved significantly. Just as successful Business Central projects rely on specialists with different skills, I believe AI works best when each agent has a focused role rather than trying to do everything.
Context and Thinking Matter More Than Intelligence
One of the biggest lessons I’ve learned is that even the most advanced AI model becomes valuable only when it understands the right context. While modern AI models possess extensive knowledge of software development, ERP systems, and business processes, they don’t automatically understand your Business Central implementation, development standards, customer processes, or the reasons behind specific customizations. Without that context, AI can produce answers that sound convincing but don’t always reflect reality. The model provides intelligence, but context provides relevance.
Equally important is teaching the agent how to think, not just what to know. Rather than jumping directly to a solution, a Business Central AI agent should investigate problems the way an experienced consultant would—understanding the issue, gathering missing information, considering standard functionality, evaluating customizations, and then recommending the next logical step. By combining the right context with a structured reasoning process, AI becomes more consistent, practical, and ultimately far more trustworthy.
Instructions and Tools Define Behavior
Another important lesson I’ve learned is that building a trustworthy AI agent is about much more than writing good prompts. While prompts influence a single conversation, well-designed instructions define how the agent behaves in every interaction. Every AI agent I build has a clearly defined role, responsibilities, reasoning process, limitations, and guidelines on when to ask for more information rather than making assumptions. These instructions create consistency, allowing the agent to behave more like an experienced Business Central consultant than a generic AI assistant.
The same principle applies to tools. Initially, I believed every agent should have access to every available capability. Over time, I realized that more tools often create more complexity. Instead, each agent should only have access to the tools it genuinely needs. A Documentation Agent requires different capabilities than a Code Review or Troubleshooting Agent. By combining clear instructions with the right set of tools, each agent becomes more focused, easier to maintain, and ultimately much more reliable.
Trust Comes from Knowing and Learning
One characteristic I admire in experienced Business Central consultants is that they don’t guess—they investigate. When they don’t have enough information, they ask questions before making recommendations. AI agents should behave the same way. Rather than making assumptions, a trustworthy AI assistant should recognize when additional context is needed. Sometimes the most valuable response isn’t an answer, but the right question. That simple behaviour builds confidence because users know the AI is reasoning carefully instead of generating convincing but unreliable recommendations.
As I built more AI agents, I also realized that trust depends on more than reasoning—it depends on knowledge. Every successful Business Central implementation creates valuable assets, including functional documentation, development standards, solution designs, support resolutions, UAT scenarios, and lessons learned. Instead of leaving this knowledge scattered across documents and conversations, it should become part of a centralized AI Knowledge Hub. By giving AI agents access to trusted organizational knowledge, they can provide recommendations that are not only intelligent but also aligned with the organization’s experience, standards, and proven practices.
The Future of Trusted AI in Business Central
After building multiple AI agents for Microsoft Dynamics 365 Business Central, I’ve become convinced that the future won’t revolve around one all-knowing AI assistant. Instead, it will be an ecosystem of specialized AI agents—each designed for a specific responsibility, whether it’s requirements gathering, documentation, development, code review, testing, troubleshooting, or operational insights. Working together and learning from the same organizational knowledge, these agents will help consultants spend less time on repetitive tasks and more time solving real business problems.
More importantly, I believe the future of AI won’t be defined by who has access to the latest Large Language Model. It will be defined by who builds the most trustworthy AI agents—agents with clear responsibilities, the right Business Central context, structured workflows, and access to organizational knowledge. In the end, the success of an AI assistant isn’t measured by how intelligent it appears, but by whether an experienced Business Central consultant is confident enough to rely on it during a real customer project.
That’s the future I’m building towards—where AI doesn’t replace consultants, but becomes a trusted member of every Business Central project team, enabling people to deliver better solutions, make better decisions, and create greater value for their customers.
Leave a Reply