Predefined questions and answers flow

Predefined question and answer flows let you guide visitors through a structured conversation. Instead of open-ended chat, the chatbot asks a series of questions, presents answer choices as clickable buttons, and responds based on the visitor's selections.

There are two ways to build such a flow in ChatLab:

• Approach 1: Conversation Flow editor (recommended) - a visual node-graph editor in Settings > Flow where you lay out the conversation as stages and transitions. The graph defines the possible stage transitions; AI interprets their conditions.
• Approach 2: prompt-based - a structured prompt in Settings > Role & Behavior that describes the questions and decision logic in plain text. Quick to set up, but the structure relies on the AI following your instructions.

In both approaches, turning on Dynamic suggested follow-ups (Settings > Chat Conversation) is what makes the answer options appear as clickable buttons in the chat.

Example

Here is a predefined flow in action. The chatbot asks about the visitor's needs, presents options as buttons, and recommends a plan based on the answers:

View the static screenshot

Approach 1: Conversation Flow editor (recommended)

The Conversation Flow editor lets you design the conversation as a visual graph. Each box is a stage with its own instructions, and the labeled arrows between stages - called transitions - decide when the conversation moves on. The current stage's instructions supplement the overall role prompt. AI interprets the transition conditions and generates the replies, so it can still misunderstand an answer or omit a question.

Here is a complete example flow built in the editor - the conversation starts at Greeting and branches into separate stages depending on what the visitor wants:

Step 1: Enable the flow

Select your chatbot and navigate to Settings > Flow. Tick the Enable Conversation Flow checkbox. The first time you open the editor it shows a ready-made sample flow - edit it or click Clear sample to start from scratch.

Step 2: Create a stage for each phase

Click Add stage for every phase of your conversation. For the advisor example above, that could be:

1. Greeting - welcomes the visitor and asks what they need help with.
2. Needs questions - asks the predefined questions one by one.
3. Recommendation - maps the collected answers to a recommendation.

Step 3: Write the stage instructions

Put the questions and answer guidance for each phase into that stage's Stage instructions. For example, the "Needs questions" stage could contain:

Ask the following questions one at a time, waiting for an answer after each:
1. What will you mainly use the chatbot for?
- Customer support
- Lead generation
- Internal knowledge base
2. Roughly how many visitors does your website get per month?
- Under 1,000
- 1,000 to 10,000
- Over 10,000
Offer the answer options after each question.

A stage with empty instructions is ignored by the flow, so fill in every stage you want the chatbot to use.

Step 4: Connect the stages with transitions

Create a transition from one stage to the next (click New connection, or drag from the dot on the right side of a stage). Give each transition a Condition - a plain-language description of when to move on, for example "Visitor said what they need help with" or "Both questions have been answered". After each visitor message the chatbot checks the current stage's transitions and advances when a condition is met.

Step 5: Mark the Start and Default stages

Make sure your greeting stage is marked START (select it and click Set as start if not). The Default stage is where the conversation returns on the next visitor message after a stage with no outgoing transitions - initially it is the same as the start stage.

Step 6: Turn on answer buttons and test

Go to Settings > Chat Conversation and enable Dynamic suggested follow-ups so the answer options render as clickable buttons. Then open the chat tester in the Overview tab and walk through the flow to verify each stage asks the right questions and the transitions fire when expected.

Keep the editor's limits in mind: a flow can have up to 15 stages, each stage up to 4 outgoing transitions, and stage instructions have a character limit that depends on your chat context size. See the Conversation Flow article for the full editor guide, including start and default stages, unreachable stages, and all limits.

Approach 2: prompt-based (lightweight alternative)

The prompt-based approach guides the conversation with instructions written directly into your chatbot's role prompt. It is a good fit when you just want to nudge the conversation through a short set of questions without setting up a full flow.

How it works

A prompt-based Q&A flow combines two features:

• Custom Role & Behavior prompt - defines the questions, answer options, and decision logic
• Dynamic Suggested Follow-ups - presents the answer options as clickable buttons after each response

When a visitor triggers the flow (for example, by asking about pricing or clicking a suggested question), the chatbot follows the instructions in your prompt. After each response, it generates follow-up buttons matching the answer choices you defined. The visitor clicks an option, and the chatbot continues to the next step.

Step 1: Write the prompt

Go to Settings > Role & Behavior and open the Custom Role Definition tab (the second mode tab, next to Predefined Role). Switching to this tab lets you replace the guided role settings with your own instructions.

The custom prompt area is split into two sub-tabs, Chat / Text and Voice Conversation. Write your flow instructions under the Chat / Text sub-tab.

Write a prompt that defines your flow. A good prompt structure includes:

1. A general role description - what the chatbot does
2. A named section - use brackets like [Section Name] to label the flow
3. Numbered questions - each with a list of answer options
4. A mapping section - links answer combinations to specific recommendations
5. A trigger instruction - tells the chatbot when to start the flow

Here is a prompt template you can adapt:

You are the assistant for [Your Company]. Help customers with [your topic].

[Advisor Flow]:
Ask the following questions one by one. After each answer, explain which option fits the user's needs.

1. [First question]?
- Option A
- Option B
- Option C

2. [Second question]?
- Option A
- Option B
- Option C

Mapping:
Option A + Option A = Recommendation 1
Option A + Option B = Recommendation 2
Option B + Option C = Recommendation 3

When user asks about [topic], proceed with the [Advisor Flow].

Step 2: Enable Dynamic Suggested Follow-ups

Go to Settings > Chat Conversation and turn on the Dynamic suggested follow-ups toggle.

This is what makes the answer options appear as clickable buttons in the chat. Without this toggle, the chatbot would list the options as text instead of interactive buttons.

Keep Hide suggested questions after first message off. Wait for the automatic Saved confirmation before testing.

Step 3: Test the flow

Open the chat tester from the Overview tab and try the conversation. Verify that the chatbot follows the correct sequence of questions and that the answer options appear as buttons.

Tips for better prompt-based flows

• Keep questions short - the chatbot generates short follow-up buttons; the exact number and wording can vary. Design your answer options to be concise.
• Use clear mappings - explicitly map answer combinations to outcomes so the chatbot gives consistent recommendations.
• Add a trigger condition - specify when the flow should start (for example, "When user asks about pricing" or "When user wants help choosing a product").
• Test variations - try different answer paths to make sure all combinations produce correct results.
• Combine with training data - the chatbot can still use your knowledge base alongside the flow, so train it with relevant content for more detailed responses.

Which approach should I use?

• Conversation Flow editor - the graph limits possible transitions and supplies the current stage's instructions alongside the overall role. Use clear conditions and test branches, as AI-generated replies and routing can still make mistakes. Best for conversations with several distinct phases. Mind the editor limits (up to 15 stages, 4 transitions per stage).
• Prompt-based - quick to set up and very flexible: everything lives in one prompt, with no stage limits. The trade-off is that the structure is only as strong as your instructions - the AI may occasionally drift from the script, especially in long conversations. Best for short, simple question sequences.

When in doubt, start with the Conversation Flow editor - it is the recommended way to build guided conversations.

Model compatibility

Dynamic Suggested Follow-ups require a compatible AI model. The feature is not available for GPT 4o, GPT 4o Mini and Gemini 3.1 Flash-Lite. If your chatbot uses one of these, switch to any other model in Settings > Model & Advanced.

Related articles

• Conversation Flow - full guide to the visual flow editor
• Dynamic suggested follow-ups - detailed guide on how follow-up buttons work
• Bot behavior customization - how to write custom role instructions