Skip to main content

A Guide to Building AI Conversational Forms

Jason Allshorn 1
Jason Allshorn
Solutions Engineer
  • Tutorial Type: Basics
  • Reading Time: 15 mins
  • Building Time: 20 mins
Chat SDK v4 2x

Swift, Kotlin, and TypeScript SDKs

Build in-app chat, calls, and live streaming

Guide to building conversational form for your chat widget

In today's fast-paced digital landscape, the integration of Artificial Intelligence (AI) has transformed the way we interact with technology, enabling smarter, more efficient, and highly personalized user experiences. One area where this transformation is most evident is in the realm of chatbots, which have evolved from rudimentary interfaces to sophisticated conversational agents capable of understanding and responding to human language with remarkable accuracy. In this tutorial, we will explore the remarkable capabilities of AI-driven functions, specifically focusing on chatbot APIs and chatbot UI for chat widget, and how they supersede traditional forms of interaction.

So, if you're eager to unlock the potential of AI-driven chatbot APIs and chatbot UI to revolutionize user experiences and take your digital interactions to the next level, you're in the right place. Let's dive into this tutorial and discover how AI is reshaping the way we interact with technology, leaving traditional forms far behind.

Learning Outcomes

  • Learn how AI-driven functions enhance user interaction, offering efficiency and improved user experience over traditional forms.

  • Acquire some skill in designing and implementing AI-driven conversational interfaces using Sendbird's OpenAI functions.

  • Gain insights into strategically deploying AI-driven interfaces on your chat widget, assessing their impact on user engagement and business processes.

Introduction: Setting the Stage

ChatGPT is changing the face of modern user interfaces to the extent that how users interact with applications is changing to become conversational.

The current most common way that users interact with an application is through HTTP requests. And HTTP requests in a client for the most part take on the shape of some kind of form. The user sets filters, inputs and/or presses buttons which trigger either the collecting of data or sending of data. But this is rapidly changing.

With Sendbird’s Open API Functions it is possible, during a natural language conversation, to gather predefined details from a user and automatically trigger the sending that data to where it needs to be. With Sendbird’s OpenAI Function integration it is possible to create in natural language scenarios to gather data then to automatically trigger the sending of the requested data in a deterministic manner.

The narrative below provides an approach to capture this transformation with the help of Sendbird’s integration with Open AI and Open AI Functions.

The Basics of OpenAI Functions in Sendbird

What Are OpenAI Functions in Sendbird?

In summary, Functions are created using natural language to listen for a particular scenario or task completion in a conversation then automatically trigger an event. At the same time also have the ability to curate or generate data to fulfill a task.

For example, to know during a conversation when a user has provided all of the required information. Then send that information to your backend.

Empowering User Experiences with Chat Widget and Conversational Form

Completing forms is a backbone of many applications, where users write information, make selections or check boxes. Application forms usually include text boxes, checkboxes, radio buttons, drop-down lists, and other elements used to gather and record information from users. However, forms are often impersonal and may not account for a user’s unique circumstance. 

Limitations of Traditional Form Filling

  1. Limited Flexibility - Difficult to caput nuanced and unexpected information.

  2. User experience - Can be tedious, off-putting and long.

  3. Inconsistency in responses - Different users may see questions differently.

  4. Accessibility issues - Forms can be challenging for people with disabilities, unless designed with accessibility in mind.

  5. Data overload - Collecting too much unnecessary data can become overwhelming and difficult to manage effectively.

Chat widget and conversational form serve as powerful tools for facilitating real-time communication and capturing user inputs in a conversational manner. By integrating AI-driven functions with chat widget and conversational form, businesses can create immersive and engaging user experiences that foster deeper connections and drive better outcomes. The revolutionary potential LLMs and Sendbird’s integrations have for user Interactions and information gathering. 

Benefits of Chat Widget and Conversational Form Integration

  • Improved User Experience: Sendbird with ChatGPT creates intuitive and human-like interactions. That is to say, users can communicate with systems in their natural language. Shifting tedious form filling into a conversation makes things easier and more comfortable, especially for those who are not tech-savvy.

  • Efficiency and Speed: Sendbird’s chatbots and virtual assistants can interact much faster than human operators. As a result quicker resolutions of issues and questions enhance efficiency.

  • Accessibility: Sendbird with LLM integration and transitioning to chat based data gathering can also assist users with disabilities, particularly those who have difficulties with traditional input methods like typing.

  • Personalization: With chat based data gathering the experience can offer personalized experiences, recommendations, clarifications, and responses. The result being services that cater for individual needs which increase user satisfaction and engagement.

  • Data Insights and Analysis: The conversations your users are having when adding data via a chat interface over a form, can be used to extract insights, trends, and patterns that are invaluable for businesses and researchers, helping in better decision-making and strategy development.

Deep Dive: Implementing AI-Driven Functions

Step 1: Conceptualize your bot.

1.2 Create an outline of the bot’s task.

Consider the target is for the bot to gather information about a job candidate in a personable, polite and professional manner.

Here is an example framework:

  1. Identify the Problem:
    1. When gathering details about a user there are issues with regard to the personal details they want to give and misunderstandings about where they are from and insensitivities around gender identification.

  2. Set Objectives:
    1. To gather information in a compassionate manner about a user’s name, their age, where they live, their nationality, and their gender identity.

  3. Map and Prioritize Tasks:
    1. User’s name and where they live are essential.

    2. Once details are gathered trigger an event to submit the requested details automatically.

  4. Visualize User Interaction:
    1. User will be talking to a bot that has the task of screening the user for a job position. One task for the bot is to gather the basic information before moving on to the next task.

  5. Decompose and Limit:
    1. In this task the bot is limited to only gathering information listed above and nothing else.

  6. Prototype and Feedback:
    1. User starts talking to the bot. The bot has the single task of gathering the required information.

  7. Test and Refine:
    1. See the testing section below.

Step 2: Test and iterate a suitable system message.

The video below demonstrates iterating a system message in Sendbird until a suitable candidate is arrived at.

The following steps achieve this goal.

  1. Create the bot and add the first system message

  2. Create a set of testing messages both positive, negative and attacking (see video).

  3. Copy and paste the messages to the Sendbird Dashboard, gather the answers and review them.

  4. Continue to build the conversation and iterate the system message until a strong production system is arrived at.

Example system message iteration:


system message

Improvement comment


Mode: Gather basic information about candidate

Task: Sequence

Gather basic information about the candidate, including their name, country, city, age, and gender.

With gender, be polite and don't push to get it if they choose not to provide it.

Prevent any other responses or diversions.

It's not clear about the country and city. Likely need to be more clear about that.


Mode: Gather basic information about candidate

Task: Sequence

Gather basic information about the candidate. Gather the following:

name, nationality, current place where they are living such as city or town, age, and gender.

With gender and age be polite and don't push to get it if they choose not to provide it.

Prevent any other responses or diversions.

Consider that the user may only give their first or last name. Also consider they might mix up about their city so check about city matching country.


Mode: Gather basic information about candidate

Task: Sequence

Gather basic information about the candidate. Gather the following:

full name, nationality, current place where they are living such as city or town, age, and gender.

With gender and age be polite and don't push to get it if they choose not to provide it.

Check where they live sounds plausible. For example that the town or city the mention is in the country they mention

Prevent any other responses or diversions.

Production Candidate

Step 3: Create an auto triggered Function

Based on the above system message v1.2 it is possible for chatGPT to guide the user to provide their details then automatically send those details to an endpoint of your choice. There is an example video below. The approach follows the following steps.

  1. Create a Function trigger prompt.

  2. Create an endpoint to send the data to.

  3. Formulate the data to be sent.

  4. Test and refine the Function trigger.

  5. Test and refine the data generation prompts.

Hypothetical Case Study: Simplifying Complex Forms with AI-Driven Chatbots in Financial Services

Consider that much in the LLM and ChatGPT is new. Therefore, the below represents a hypothetical case study to emphasize the capabilities of Sendbird with Open AI Functions.


"MoneyMatters," a fictitious financial services firm, encountered low customer engagement and high drop-off rates due to their lengthy and complex form-based processes for financial applications.


To replace intricate form-filling processes with a user-friendly, conversational AI-driven chatbot system.


  • Integration: MoneyMatters deployed Sendbird's AI chatbots to guide customers through their financial application processes in a conversational format.


  • Form Complexity: Traditional forms contained complex, industry-specific jargon that often confused customers.

  • User Navigation Issues: Customers frequently got lost in the multi-step form process, leading to incomplete applications.


  • Simplified Conversations: The chatbot was programmed to break down complex financial concepts into simple, conversational language.

  • Guided Process: The chatbot led customers step-by-step, ensuring they provided all necessary information without feeling overwhelmed.


  • Improved Completion Rates: Application completion rates soared due to the streamlined, guided process.

  • Enhanced User Experience: Customers reported a more engaging and less stressful application process.

  • Operational Efficiency: The firm observed a significant reduction in customer service inquiries related to form completion.

Fictitious Conclusion

  • By replacing complex forms with an AI-driven chatbot, MoneyMatters enhanced the efficiency and user-friendliness of their financial application process, leading to increased customer satisfaction and higher completion rates.

Testing and Refinement

One key to successful implementation is iterating and refining the process outlined above. Recording outcomes and creating testing scenarios takes time. Consider using chatGPT’s standalone interface to assist with creating testing examples. Also, take time to share with your internal users and get their feedback. Additionally, provide end users with the capability to also provide feedback. The journey with Sendbird and ChatGPT is grounded in the requirement of continuous improvement. Therefore, try it out, monitor the output, iterate and improve.

Conclusion: Revolutionizing User Experiences with AI-Driven Functions, Chat Widget, and Conversational Form

In wrapping up, we can see there is a shift from traditional forms to AI-driven interactions, particularly with Sendbird and ChatGPT. The concept of moving from forms to conversations, marks a notable evolution in how we interact with technology. This move towards more conversational, AI-powered interfaces isn't just about keeping up with trends; it's a game-changer in enhancing user engagement and streamlining data management.

As we delve deeper into this new realm, the potential for more personalized, efficient digital experiences is immense, signaling a promising direction for the future of user interfaces.

At Sendbird, we specialize in developing customized solutions that leverage AI-driven functions, chat widget, and conversational forms to help businesses unlock new possibilities for user engagement and satisfaction. Contact us today to learn how we can empower your business with advanced AI-driven capabilities tailored to your specific needs and objectives.