Configuring General Instructions & Persona

General Instructions define who your AI Agent is and how it behaves in every conversation. Think of them as the agent's personality and code of conduct — always active, regardless of which Policy is triggered or which Document is retrieved.

This is where you give the agent its name, tone, language rules, and any global boundaries it must always respect.

General Instructions vs. Policies and Documents

It helps to understand how General Instructions relate to the other two training tools:

• General Instructions are always active. They define the agent's baseline identity and behavior across every single conversation.

• Documents give the agent knowledge to draw on when answering general questions.

• Policies override everything else when triggered — they tell the agent exactly what to do in a specific situation, step by step.

A good mental model: General Instructions define who the agent is. Policies define what it does in specific situations. Documents define what it knows.

What to include in General Instructions

Persona and identity

Give the agent a name and a role. This sets the tone for every interaction and makes the agent feel consistent and on-brand.

Example: "You are Sofia, a customer support agent for Engaige. You represent the brand with warmth and professionalism."

Tone and communication style

Describe how the agent should write. Be specific — "friendly" means different things to different brands.

Example: "Write in a warm, conversational tone. Keep responses concise. Avoid corporate jargon. Never use overly formal language like 'Dear Sir/Madam'."

Language rules

Define how the agent handles language. The most common rule is to always match the customer's language.

Example: "Always respond in the same language the customer uses. If ensure, default to english."

Scope and topic boundaries

Tell the agent what it handles and what it doesn't. This prevents it from going off-topic or making commitments it shouldn't.

Example: "You handle order inquiries, returns, delivery questions, and product information. For billing disputes or account deletions, always transfer the conversation to a human agent."

Brand rules and hard limits

Add any non-negotiables: things the agent must always or never do, regardless of context.

Examples:

• "Never mention competitors or compare our products to theirs."

• "Never promise a specific delivery date unless you have retrieved the tracking information."

• "Always refer to our return window as '30 days from delivery', not '30 days from purchase'."

Escalation defaults

You don't need to specify escalation rules here if you're already handling them in Policies: but if you want a global fallback, you can add it.

Example: "If a customer expresses strong frustration or explicitly asks for a human, transfer the conversation immediately without trying to resolve it first."

How to configure General Instructions

1. Go to AI Agents and open the agent you want to configure.

2. Navigate to the Behavior tab.

3. Find the General Instructions field at the top of the page.

4. Write your instructions in plain text. No special formatting is required — the agent reads them as natural language.

5. Save your changes.

Tips

• Write instructions as directives, not descriptions. "Always respond in French" works better than "The agent speaks French."

• Keep it focused on behavior, not knowledge. Product details, FAQs, and policies belong in Documents or Policies — not here. General Instructions should stay short and behavioral.

• Don't repeat what Policies already handle. If you have a Policy for cancellations, you don't need to explain cancellation behavior in General Instructions. Reserve this space for things that apply to every conversation.

• Test in the Playground after making changes. Trigger a few different conversation types to check that the tone and rules feel right across scenarios.