Best Practices

Get the most out of your AI assistant by following these proven strategies.

The Two-Layer Approach

For best results, use both approaches together:

1. Upload Documents (Knowledge Base)

Upload your manuals, FAQs, and policies as PDFs in the Documents page.

• Great for general Q&A
• AI searches through content for relevant info
• Not enforced as strict rules

2. Configure AI Rules (Strict Enforcement)

Add critical policies in the AI Rules page.

• Enforced every time
• Custom terminology always applied
• Guardrails prevent costly mistakes

Example: Event Ticketing Platform

Step 1: Upload Manual

Upload your support manual PDF in Documents for general context.

Step 2: Add Terminology

In AI Rules → Terminology, add your custom terms:

Instead of	Use
ticket	Pass
purchased	Locked In
event group chat	Party
user group	Squad

Step 3: Set Policies

In AI Rules → Policies:

Refund Policy:

Non-refundable: VAT (7.5%), Platform Fee (5%), Gateway Fee.
Full refund only for event cancellation or host negligence.

Escalation Rules:

Escalate to human if:
- Transaction amount > ₦100,000
- 3rd failed resolution attempt
- Customer mentions lawyer or social media threats

Step 4: Configure Guardrails

In AI Rules → Guardrails:

• Max Auto-Refund: 100000
• Auto-Escalate Keywords: lawyer, twitter, court, ridiculous

Quick Tips

Terminology

Add any brand-specific terms customers see in your app. This keeps AI responses consistent with your product.

Guardrails

Set conservative limits initially. You can always increase auto-approve thresholds as you gain confidence.

Escalation Keywords

Include words that indicate frustration or legal threats. Better to escalate early than risk a bad customer experience.