community postCopilot & Agentscommunity - not Microsoftreviewed - checked by Opus

Agents in Business Central – part 2 – the prompt

Building effective agents in Business Central requires well-structured prompts with clear responsibilities, guidelines, and task instructions. The post explains how to use memory, page-specific instructions, and conditional logic to guide agent behavior reliably and keep humans in control.

post Discover Microsoft Business Central - The wonderful world of Business Central

Bert Verbeek · 2026-02-16 · 722 words · about 3 min

"The way you guide and structure your agent's behavior determines whether it becomes a reliable assistant or a confused chatbot."

Key points

  • Agent instructions need three core components: responsibilities (scope and boundaries), guidelines (rulebook for behavior), and task instructions (concrete procedural steps)
  • Agents should ask for help when data is missing or unclear, and request user review for important steps to maintain human oversight
  • Agents can memorize specific actions, search results, and explicitly instructed values, but not the full state of every page
  • Use page-specific instructions with conditional logic to provide detailed guidance only when needed and improve performance
  • Business Central agents have context-aware tools like searching pages, selecting rows, executing actions, and setting field values

Quotes

  • “The way you guide and structure your agent’s behavior determines whether it becomes a reliable assistant or a confused chatbot.” (Highlights that prompt quality directly impacts agent reliability and usefulness)
  • “This memory model protects performance but still allows context continuity.” (Explains the practical balance between agent capability and system performance)

Context

  • Features: Agent prompting framework, Responsibility definition, Guidelines and rules, Task instructions, Memory management, Page-specific instructions, Conditional logic, Agent Tasks action with View Log Entries

Source: Discover MS BC - Bert Verbeek, community blog. Summary, key points and quotes are derived (CONTENT-NOTICE.md); read the original for the full text.