Setting up AI Chatbot ( AI Assistants )

FlowMattic now supports creating powerful AI Chatbots powered by OpenAI. You can create assistants tailored to your business needs, upload relevant content to train them, and embed or share them easily.

Before proceeding, ensure you’ve completed the following prerequisite.

🔧 Prerequisite: Connect Your OpenAI Account

Before setting up your AI Assistant, you must connect your OpenAI account to FlowMattic using Connects.

Follow this step-by-step guide:
👉 Add OpenAI API Key to FlowMattic Using Connects

This connection is required so FlowMattic can send and receive data from OpenAI’s API using your account.

🧠 Step 1: Navigate to AI Assistants

Click the AI Assistants tab from the sidebar (see Screenshot 1, marked as 1).

If you haven’t created any assistants yet, you’ll see an empty list. Click the Create New Assistant button (Screenshot 1, marked as 2).

➕ Step 2: Create a New Assistant

A popup will appear where you can enter the assistant details:

• Assistant Name: Choose a unique name.
• OpenAI Connect: Select the OpenAI connection created earlier.
• Assistant Description: A short description for internal reference.

Once filled, click Create.

⚙️ Step 3: Configure Your AI Assistant

Once your assistant is created, it will appear in the list. Click the Edit button to configure it further.

You’ll be redirected to the AI Assistant Setup Page, which includes the following tabs:

🔹 Setup Tab

• Chatbot Name

• Choose Connect – Select the OpenAI Connect you created.

• Description – Internal reference text.

• Choose Assistant – From your OpenAI Assistant list or create a new one.
  Click the Create Assistant button to open an advanced assistant creation modal (Screenshot 5).

  You can configure:

  • Assistant Name
  • Connect
  • Model (e.g., gpt-4-turbo)
  • System Instructions
  • Description

• Welcome Text – The message users will see first.

• User Input Placeholder – The hint in the input field.

• Instructions – System-level instructions for your assistant.

• AI Model – Select model (make sure it’s compatible with retrieval if you use content sources).

• Load User Session Chats – Yes to retain session history for users.

Click Save Changes once done.

📂 Step 4: Add Content Sources (Training Files)

Click the Content tab > then click Add Source (See Screenshot 6).

You’ll see the Add Content Source modal where you can:

• Upload a content file that your assistant will use to answer questions.
• Add a short description for the source.

📁 Supported File Types

• .txt, .doc, .docx, .pdf, .html, .json, .md
• Max file size: 10 MB

Once uploaded, these files are sent to OpenAI and used by the assistant to answer user queries accurately. Make sure to upload content relevant to your assistant’s role (e.g., FAQs, help docs, policy files).

Note: Ensure you are using a model that supports retrieval when working with file-based content.

🎨 Style Tab – Customize Your Chatbot’s Appearance

The Style tab allows you to customize the visual appearance of your AI assistant to match your brand or website aesthetic. You can adjust:

• Chat Window Header Title – Set the chatbot’s title shown at the top of the chat.
• Theme Color – Choose a primary color to apply to the header and button elements.
• Chat Bubble Color – Customize the color of chat bubbles for both user and assistant.
• Font Family – Select from system fonts for consistency with your site typography.
• Position on Screen – Decide where the chatbot will appear (bottom left/right, full page).
• Icon and Branding – Upload your custom logo or icon to personalize the experience.
• Chat Button Text and Icon – If the assistant appears minimized, customize the button text and choose a recognizable icon for your brand.

📝 Tip: Preview your assistant style in real-time as you customize, so you get the exact look before embedding.

⚡ Actions Tab – Add Dynamic Workflows to Your Assistant

The Actions tab allows you to go beyond answering questions. Here, you can configure the assistant to trigger real workflows or perform tasks based on specific conditions or phrases.

For example:

• 🔁 Trigger a FlowMattic workflow when the user asks for a report, submits feedback, or wants to schedule something.
• 🔗 Pass dynamic input to workflows (e.g., user’s question or a selected option).
• 🔄 Update records, send emails, or perform API calls based on user input.

Actions Configuration Includes:

• Trigger Word/Phrase – Define what user input should trigger an action.
• Mapped Workflow – Choose an existing FlowMattic workflow to run.
• Pass Input Variables – Use dynamic fields to send data into the workflow like {{user_input}}.

🔍 Example: If user says “Generate invoice”, it can run a workflow that collects data, generates a PDF invoice, and sends it via email.

💬 Conversations Tab – Monitor and Manage User Chats

The Conversations tab displays a complete history of all user interactions with your assistant. It’s useful for reviewing:

• ✅ Successful responses – See how the assistant handled questions.
• ⚠️ Unanswered queries – Identify gaps where training or content can be improved.
• 📁 Content usage – Understand which documents or instructions the assistant referred to.
• 🧑‍💻 User sessions – View interactions grouped by session and user, making it easy to trace back any conversation.

💡 This tab is essential for refining your assistant over time by observing real-world usage and feedback.

🔗 Embed or Share Your Assistant

Once your AI Assistant is ready, you can either embed it on your website or share it via link. These options are located at the top right corner of the setup interface.

🔧 Embed Assistant

• Click the Embed button to get a code snippet.
• You can copy and paste this script into your website’s HTML where you want the chatbot to appear.
• The embedded assistant will retain all configurations including style, content, and memory.

✅ Use this if you want a persistent chatbot on your landing pages, help centers, or member dashboards.

🔗 Share Assistant (Link)

• Click the Share button to generate a public URL.
• This link can be shared directly in emails, support docs, or social media.
• The assistant will load in a standalone view with full chat interface.

📣 Ideal for sending quick access to clients, prospects, or internal teams without embedding it on a site.

🧠 Real-World Examples of AI Assistants

Here are some powerful ways to use AI Assistants in FlowMattic, particularly with uploaded content and workflow actions.

🔹 Example 1: Support Assistant ✅

Use Case: Provide automated customer support using your company’s help docs.

• Content Sources: Upload refund policies, setup guides, onboarding manuals, and FAQs.
• Actions:
  • Escalate to human support via a FlowMattic workflow when needed.
  • Generate a support ticket with the chat transcript and user email.

🔹 Example 2: Documentation Help Assistant 📚

Use Case: Help users understand your technical product or API.

• Content Sources: Upload your developer docs, API references, and integration tutorials (in .pdf, .md, .html, or .json).
• Instructions: Ask the assistant to respond in a technical tone with examples when available.
• Perfect For: SaaS platforms, API providers, plugin documentation portals.

🔹 Example 3: HR Assistant for Company Policies 🧑‍💼

Use Case: Internal AI assistant for employees to ask about HR and company policies.

• Content Sources: Upload your company handbook, HR policy, leave rules, holiday calendar, etc.
• Instructions: Answer based on company rules and direct employees to the right department when needed.
• Bonus: Use actions to send leave applications to HR or log attendance queries.

🔹 Example 4: Product Assistant for eCommerce ✅

Use Case: Assist customers with product-related queries directly from your catalog.

• Content Sources: Upload product catalogs, pricing tables, warranty documents, and usage manuals.
• Actions:
  • Suggest products based on user needs.
  • Generate a discount code or cart link using a FlowMattic workflow.
  • Send availability updates to users’ emails.

🔹 Example 5: Knowledge Base Assistant for Teams 🧠

Use Case: Enable internal teams to quickly search and reference internal training or SOPs.

• Content Sources: Upload SOPs, onboarding files, technical how-tos, and internal checklists.
• Use Case: Instead of digging through folders, employees just ask the assistant.
• Tip: Create multiple assistants for different departments (Sales, Support, DevOps).

🔹 Example 6: Client Onboarding Assistant 👥

Use Case: Guide new clients or partners through your onboarding process.

• Content Sources: Upload welcome decks, implementation checklists, and success guides.
• Instructions: Provide friendly and professional guidance.
• Actions:
  • Collect client details using input.

❓ Frequently Asked Questions (FAQs)

🔸 Is it safe to upload sensitive content to the assistant?

Yes. All uploaded files are securely handled and sent to OpenAI’s servers only for use within your assistant. However, avoid uploading files with sensitive credentials or PII unless your OpenAI usage policy covers such data securely.

🔸 Can I use multiple assistants for different websites or use cases?

Absolutely! You can create unlimited assistants in FlowMattic and configure each with unique styles, content, actions, and embed code.

🔸 Does the assistant remember past conversations?

Yes, if you enable “Load User Session Chats” under the Setup tab. This keeps chat history for users during a session so they can continue conversations without losing context.

🔸 Can I update the content files later?

Yes. You can revisit the Content tab, remove old files, and upload new ones anytime. The assistant will start using the updated content immediately once uploaded and indexed.

🔸 Which OpenAI models are supported?

You can use any model available in your OpenAI account such as gpt-3.5-turbo, gpt-4, or gpt-4-turbo. If you’re using content sources (files), make sure the selected model supports retrieval (like gpt-4-turbo with files).

🔸 How can I test the chatbot before embedding?

Use the preview inside the configuration page. Type in test queries and verify if responses are accurate and actions trigger as expected. You can always click Restart to reset the chat session.