Chat Agent Overview

Chat Agent Overview

This article explains how to build an AI bot with Flora and quickly integrate it into your business. The key to a successful bot is a well-crafted base system prompt.

What Is a Base System Prompt?

The base system prompt is the core instruction set that tells your Chat Agent:

• How to act (its behavior and style)

• How to respond (tone, level of detail, format)

• What to focus on (goals, priorities, constraints)

• How to think (its internal reasoning and chain of thought)

• How to interact with users (what it can and cannot do)

Every message your bot sends is guided by this base system prompt, so investing time in writing it well is crucial.

Why the Base System Prompt Matters

A strong base system prompt helps you:

• Align the bot with your business
  Ensure the bot reflects your brand voice, policies, and domain knowledge.

• Control user experience
  Define how the bot greets users, asks questions, and handles confusion or errors.

• Guide reasoning and consistency
  Make the bot follow consistent steps when answering, troubleshooting, or making recommendations.

• Reduce configuration overhead
  Many behaviors can be controlled in one place (the prompt) instead of scattered settings.

Key Elements of a Base System Prompt

When creating your base system prompt in Flora, consider including the following elements:

1. Personality

Define how the bot should “sound” and behave:

• Tone (e.g., professional, friendly, concise, technical)

• Level of formality

• Use of jargon or plain language

• Whether it uses humor or stays strictly factual

Example:

You are a friendly, professional support assistant for [Company Name]. You write in clear, concise sentences, avoid slang, and never use emojis. You are patient and respectful, even when users are frustrated.

2. Role

Clarify what the bot is and what it is responsible for:

• Its position (e.g., “customer support agent”, “sales assistant”, “HR helpdesk bot”)

• Its scope (what it covers and what it does not)

Example:

Your role is to act as a first-line customer support agent for our SaaS platform. You help users with account issues, billing questions, and basic product troubleshooting. For legal or security questions, you must escalate to a human agent.

3. Goal

Specify what the bot should optimize for in each interaction:

• Primary objectives (e.g., resolve issues, qualify leads, educate users)

• Secondary objectives (e.g., collect feedback, suggest upgrades)

Example:

Your primary goal is to resolve the user’s issue as efficiently as possible while maintaining clarity and accuracy. When appropriate, you may suggest relevant features or upgrades that genuinely help the user.

4. Chain of Thought (Reasoning Process)

Define how the bot should think and structure its responses. This is where you control the reasoning steps the bot follows.

You can instruct the bot to:

• Ask clarifying questions before answering

• Follow a step-by-step troubleshooting process

• Summarize the user’s problem before proposing a solution

• Provide answers in a specific structure (e.g., “Summary → Steps → Notes”)

Example:

Before answering, always:

Briefly restate the user’s question in your own words to confirm understanding. If information is missing, ask up to 2 concise clarifying questions. Provide a step-by-step solution. End with a short summary or next steps.

You can also specify how much of this reasoning should be visible to the user (e.g., keep internal reasoning concise and user-facing explanations clear and simple).

5. Interaction Rules

Set boundaries and guidelines for how the bot interacts with users:

• What it must not do (e.g., no medical, legal, or financial advice)

• When to escalate to a human

• How to handle unknowns or missing data

• How to respond to inappropriate or off-topic messages

Example:

If you are not confident in an answer or lack information, say so clearly and suggest contacting human support.
If a user asks for medical, legal, or financial advice, politely decline and recommend speaking to a qualified professional.
If a user is abusive, remain calm, do not respond in kind, and suggest ending the conversation or contacting support.

6. Content and Knowledge Scope

Explain what information the bot should rely on:

• Internal knowledge bases or documents

• Product documentation

• Policies and FAQs

• What to do when information is not available

Example:

Use the company knowledge base, product documentation, and FAQs as your primary sources of truth. If the requested information is not available in these sources, say that you do not have that information and offer alternative ways the user can get help.

Example Base System Prompt Template

You can adapt this template when creating your Chat Agent in Flora:

You are [Bot Name], an AI chat agent for [Company Name].

Personality:

Friendly, professional, and concise. Use clear, simple language and avoid technical jargon unless the user shows they are technical. No emojis or slang.

Role:

Act as a first-line support and information assistant for [describe products/services]. Help users with [list main topics: e.g., account setup, billing, product usage, troubleshooting].

Goals:

Resolve user issues accurately and efficiently. Keep responses short but complete. When relevant, suggest helpful features or resources without being pushy.

Chain of Thought / Process:

Restate the user’s question in 1 short sentence to confirm understanding. Ask for missing details if needed (maximum 2 questions at a time). Provide a clear, step-by-step answer or solution. If the issue cannot be fully resolved, explain what you can and suggest next steps or escalation.

Interaction Rules:

If you are unsure or lack information, say so clearly and avoid guessing. Do not provide medical, legal, or financial advice. For security or account-specific issues you cannot verify, instruct the user to contact human support via [channel].

Knowledge Scope:

Use the provided company documentation, FAQs, and policies as your main sources. If the answer is not in these sources, say that you do not have that information and suggest contacting support.

Integrating the Chat Agent into Your Business

Once your base system prompt is ready in Flora, you can integrate the Chat Agent into your existing workflows:

1. Configure the Agent in Flora

• Create a new Chat Agent.

• Paste and customize your base system prompt.

• Connect relevant knowledge sources (docs, FAQs, policies).

1. Choose Integration Channels

• Website widget (for customer-facing support)

• In-app chat (for product guidance)

• Internal tools (for employee helpdesk or internal FAQs)

• Other channels supported by Flora (e.g., CRM, helpdesk platforms)

1. Test and Refine

• Run sample conversations that reflect real user scenarios.

• Adjust personality, role, and rules in the base system prompt based on results.

• Iterate until the bot’s behavior consistently matches your expectations.

Best Practices

• Start specific, not generic: clearly define role, scope, and limits.

• Keep the prompt organized with headings or bullet points.

• Review and update the prompt regularly as your product or policies change.

• Use real user conversations to identify gaps and refine instructions.

• Document your final base system prompt so your team understands how the bot is configured.

A well-designed base system prompt is the foundation of an effective Chat Agent in Flora. By clearly defining personality, role, goals, reasoning, and interaction rules, you can quickly deploy a bot that fits seamlessly into your business and delivers consistent, high-quality support.

.