AI Persona & Knowledge Base

Configure how Moxabot thinks, talks, and responds. This guide covers building your AI persona, populating your Knowledge Base, testing replies in the Preview tab, and setting up hand-off rules that escalate conversations to your human team at exactly the right moment.

Overview

Moxabot's AI is powered by Google Gemini 2.5 Flash — a fast, highly capable model optimised for conversational support. Every reply it generates is shaped by three things you control:

Set up your AI Persona

The AI Persona page is where you give Moxabot its identity. Everything you set here influences the tone, style, and behaviour of every AI-generated reply across all your connected channels.

Persona name

Give your AI a name that fits your brand — "Aanya", "Max", "Support Bot", or anything else. This name appears in the chat widget header, in the greeting message, and in agent-facing inbox notes when the AI replies. Keep it short (1–2 words) and human-sounding for best results.

Tone

Choose one of three built-in tones to set the overall register of every AI reply:

Greeting message

The greeting message is the first thing a visitor sees when they open the chat widget or start a new conversation on any channel. Write 1–2 sentences that:

• Introduce your AI by name
• Set expectations ("I can help with orders, returns, and product questions")
• Invite an action ("How can I help you today?")

You can include an emoji if it matches your brand tone. Keep it under 160 characters so it renders cleanly on SMS channels.

Instructions

Instructions are a free-text field where you give the AI behavioural rules — things the AI should always do, never do, or do only in specific situations. Think of it as a one-page brief for a new customer service hire.

Language

Moxabot auto-detects the language the customer is writing in and replies in the same language by default. To restrict replies to a specific language, add a line to your Instructions such as: "Always reply in English, regardless of what language the customer uses."

Build your Knowledge Base

The Knowledge Base is where you teach the AI about your business. Every time the AI generates a reply, it searches your Knowledge Base for relevant content and uses it to answer accurately. The more relevant content you add, the better — and more confident — the AI's answers become.

Adding content

Navigate to Knowledge in the sidebar. You have three ways to add content:

Best practices

You don't need to add everything at once. Here's a recommended starting point:

Writing good Knowledge Base articles

When writing articles directly in Moxabot, use a question-and-answer format. Each article should:

• Have a clear, specific title phrased as the question customers actually ask ("How do I return an item?")
• Open with a direct answer in the first sentence
• Include any relevant URLs, timelines, or policy details in the body
• Stay focused — one topic per article is better than one long article covering many topics
• Avoid internal jargon the customer wouldn't use

Good article title:  "How long does standard shipping take?"
Bad article title:   "Shipping"

Good opening:  "Standard shipping takes 3–5 business days for orders placed before 2pm."
Bad opening:   "There are several shipping options available to customers."

Good body: Includes the cut-off time, weekend handling note, and a link to the tracking page.
Bad body:  Vague — "It depends on your location and order date."

Test & preview AI replies

Before you go live, use the Preview tab to test your AI in a sandboxed environment. Replies in Preview are never delivered to real customers and don't count toward your conversation limits.

Common issues and fixes

Configure hand-off rules

Hand-off rules define the conditions under which Moxabot stops trying to resolve a conversation automatically and assigns it to a human agent. Getting this right is critical — you want the AI to handle volume, but never leave a frustrated customer talking to a bot when they need a person.

Trigger phrases

Trigger phrases are keywords or short phrases that, when detected in a customer's message, immediately escalate the conversation. Navigate to AI Persona → Hand-off and add your triggers one per line. Common examples:

Escalation flow

When a trigger phrase is detected, here is exactly what happens:

What happens when no agent is available

If a hand-off is triggered outside of your team's working hours or when all agents are busy, the AI sends the hand-off message and the conversation enters the escalation queue. The customer is not left hanging — the AI also sends a follow-up message configured in Business Hours settings(e.g. "Our team is currently offline — we'll get back to you within 4 business hours.").

Tips for writing great instructions and KB articles

Limits

Next steps