What is a Form Assistant suitable for?
A Form Assistant helps you collect Member Properties through a conversation rather than using a traditional form.
This creates a natural dialogue in which the chatbot gathers the required information step by step and stores it in a structured way.
The Form Assistant is particularly suitable when Users:
should be guided through an input step by step
need support formulating their answers
should work out information together with the bot
This makes it much easier to collect more complex information or information that requires explanation.
Set up the Form Assistant
To create a Form Assistant, navigate to the Prompt section in the editor. There, select the Form Assistant type in the Personality settings. You will then be prompted to add a Member Property.
If you have not yet created any Member Properties, first take a look at the relevant Intercom article.
Once a Member Property has been added, define two prompts:
the Task Prompt
the Definition for achieved property
These two settings determine what the bot should achieve and when the goal is considered fulfilled.
The Task Prompt
The Task Prompt works similarly to a traditional chatbot prompt.
Here, you define:
what goal the bot pursues
how it should ask for the information
how it interacts with the User
The more clearly you formulate the Task Prompt, the more targeted the Form Assistant will work.
Example
A fitness coach wants to find out which sport a Member participates in:
“You are a friendly fitness assistant. Your task is to find out which sport the User primarily participates in. Ask targeted follow-up questions if the answer is unclear or too general. Respond in a motivating and easy-to-understand way. Once the sport has been clearly named, work toward a clear conclusion.”
Definition for achieved property
Here, you define when a Member Property is considered complete.
This means that you specify when the bot recognizes the information as sufficient and saves it.
Example:
“The property is achieved when the User has clearly stated which sport they primarily participate in. Save only the specific sport, e.g. ‘Running’, ‘Football’, or ‘Yoga’.”
This way, the bot knows exactly what it should work toward.
Advanced workflows
The Form Assistant also makes it possible to combine multiple Member Properties to map more complex, multi-step processes.
Instead of collecting just one piece of information, the bot can determine several related data points step by step.
Example
You can combine multiple Member Properties to create multi-step processes.
The Form Assistant guides the User through a connected conversation.
Example:
A fitness coach collects the following one after another:
Sport → Which sport does the User participate in?
Problem or challenge → What is the User currently having difficulties with?
Previous attempts at solving it → What has the User already tried?
The bot builds on the previous answers and guides the User through a logical process.
Benefits
With the Form Assistant:
complex information is collected in a structured way
each answer builds on the previous one
a significantly better User experience is created
All answers are automatically saved in the respective Member Properties.
This allows you to:
analyze data in a targeted way
use it in Automations
keep a quick overview of your Members
User experience while completing it
For Members, the Form Assistant feels like a normal conversation.
As soon as a property is recognized, the chatbot displays this transparently and suggests saving the answer.
Users can:
review the recognized answer
adjust it
or save it directly and continue
This means they retain control over their information at all times.
Note on model selection
The choice of AI Model affects the quality of the Form Assistant.
Smaller models (e.g. Mini or Nano) can lead to less consistent results. The bot may ask multiple times even though the answer has already been provided.
For reliable use, we recommend:
powerful base models (e.g. GPT models)
alternatively, Gemini models (Flash versions also work reliably)
Choosing the right model ensures that inputs are recognized correctly and processed efficiently.
If you have any questions, please do not hesitate to contact our support team via the HelpCenter in the bottom-left corner. We usually respond within 5 minutes on weekdays from 09:00 to 17:00.