Appending Custom System Prompts

Appending custom system prompts allows you to dynamically modify how a chatbot behaves during a conversation, without changing its base persona. This is a workflow-level control mechanism used to inject temporary or conditional instructions into the AI’s system context.

This is one of the most powerful chatbot controls in Aimogen.

---

What “Appending a System Prompt” Means #

Every chatbot has a base system prompt, defined by its persona (role, behavior, tone).

Appending a custom system prompt means:

• adding extra system-level instructions
• at runtime
• based on triggers or conditions
• for part or all of a conversation

These instructions influence how the AI responds, not what the UI does.

---

How This Differs from Persona Prompts #

Persona prompt:

• always active
• defines the chatbot’s identity
• changes behavior globally for that chatbot

Appended system prompt:

• added conditionally
• temporary or scoped
• layered on top of the persona
• can be removed or overridden later

Think of it as a runtime behavior modifier, not a new persona.

---

Where Appended Prompts Are Used #

Custom system prompts are appended via:

• chatbot triggers
• hardcoded message workflows
• conditional actions

They are not typed by users and not visible in the UI.

---

When Appended Prompts Are Evaluated #

Appended system prompts are applied:

• when a trigger fires
• before the AI generates its next response
• alongside the base persona and conversation context

They affect subsequent AI replies, not past messages.

---

Common Use Cases #

Appending system prompts is ideal for enforcing rules such as:

• “Only answer questions about pricing.”
• “Do not provide technical advice.”
• “Speak very briefly and ask follow-up questions.”
• “Focus on upselling premium plans.”
• “Answer as a support agent, not as a sales assistant.”
• “Avoid collecting personal data.”

These rules would be risky to rely on AI judgment alone.

---

Scoped vs Persistent Prompt Appending #

Depending on configuration, appended prompts can be:

• One-response only
  Applies to the next AI message, then discarded.
• Conversation-scoped
  Remains active for the rest of the conversation.
• Until condition changes
  Removed or replaced when another trigger fires.

Scope control is critical to avoid unintended long-term behavior changes.

---

How Appended Prompts Interact with AI #

System-level instructions:

• override user intent if conflicting
• take priority over user messages
• strongly influence tone, content, and boundaries

If the appended prompt conflicts with the persona:

• the most recent system instruction usually wins

This makes appending powerful, but also dangerous if misused.

---

Appending vs Hardcoded Messages #

Hardcoded messages:

• send fixed text
• do not involve AI generation
• cost no tokens

Appended system prompts:

• still generate AI responses
• influence how responses are generated
• consume tokens
• affect reasoning and tone

Both are often used together in workflows.

---

Examples of Workflow Patterns #

Lead qualification flow

• append: “Ask for email before continuing.”
• AI asks naturally
• remove prompt after email is collected

Escalation flow

• append: “Respond as a senior support specialist.”
• AI adopts stricter tone
• persists until conversation ends

Compliance enforcement

• append: “Do not give legal or medical advice.”
• applies to all future responses

---

What Appended Prompts Do Not Do #

They do not:

• change the model
• retrain the AI
• affect other chatbots
• modify placement rules
• alter UI or styling
• override consent gating

They are purely behavioral instructions.

---

Performance and Cost #

Appending prompts:

• increases prompt size
• slightly increases token usage
• does not add extra API calls

Poorly scoped prompts can increase cost over long conversations.

---

Common Mistakes #

• appending overly long instructions
• stacking multiple conflicting prompts
• forgetting to remove or override prompts
• using prompts instead of workflows for logic
• encoding business rules in natural language instead of conditions

Prompts should guide behavior, not replace logic.

---

Best Practices #

• keep appended prompts short and explicit
• document why each prompt exists
• scope prompts as narrowly as possible
• prefer one-purpose prompts
• combine with conditional logic
• test prompts in the backend Playground first

Treat appended prompts like runtime configuration flags.

---

Summary #

Appending custom system prompts allows Aimogen chatbots to change behavior dynamically during conversations, without altering their core persona. Applied through workflows and triggers, these prompts act as high-priority behavioral instructions that guide AI responses deterministically. Used carefully, they provide precise control over tone, scope, and boundaries; used carelessly, they can override intent and cause unpredictable results.