Why should you care?

Hospital and ERP assistants must produce human-readable answers, not just scores, so generative models sit at the center of AgentNest copilots.

See it live — copy this example

Paste into an ASP.NET Core 8+ / AgentNest project, then run with dotnet run (set your API keys in user-secrets).

// AgentNest.Agents/Prompts/EmailDraftPrompt.cs
public static class EmailDraftPrompt
{
    public const string System = """
        You draft concise B2B follow-up emails for AgentNest CRM.
        Never invent pricing. Use only facts in the user message.
        """;
    public static string User(string leadName, string lastMeeting) =>
        $"Draft a follow-up to {leadName} referencing {lastMeeting}.";
}

What happened?

• System and user prompt strings separate policy from per-request facts.
• CRM controllers pass lead context into User() before calling a chat client.

Practice next

1. Add AgentNest.Agents class library to the solution.
2. Create EmailDraftPrompt with system rules and a User builder.
3. Register prompts as singletons or static helpers — no secrets inside strings.
4. Add a MaxWords rule to the system prompt and test truncation logic.
5. Build a User() overload that accepts bullet facts from a DTO.

Remember

Generative AI in AgentNest turns structured CRM/ERP data into drafts. Separate system policy from user facts in prompt builders. Always plan a human or rules review before sending generated content.

CRM follow-up drafts

Reps want one-click follow-ups after discovery calls in AgentNest CRM.

Outcome: Prompt templates produce drafts; reps edit before send, cutting write time by half.