top of page
Meghana Alavala Logo Horizontal - Color Mix 2.png

Implementing UX in AI Chatbot

Introduction

In today's fast-paced digital world, users expect instant responses when interacting with customer service platforms. DaxP, a healthcare service provider, recognized this need and aimed to design an AI chatbot, Ask DaxP, that delivers seamless and efficient user experiences. To achieve this, we explored best practices in chatbot UX, learning from established studies and case reports on chat interactions.

The Need for Speed in Customer Interactions

In a study done by Julia Xavier on Live chat for customer service, it is recognised that

The average wait time, in terms of how long it takes for a company to acknowledge a live chat message, is 2 minutes and 40 seconds. That’s almost as long as average wait times for phone support. Customers contacting via email expect a response within 6 hours, whereas chat customers expect an immediate response, in less than 30 seconds.

But a customer support chatbot can assist users 24/7. This reduces company costs and also saves a lot of time for real agents. This way, they only need to handle more complex inquiries that a bot can’t answer.

Designing Chat Entry Points: Placement and Visibility

Research from the Nielsen Norman Group (NNG) on The user experience of customer service chat highlights that users often ignore floating action buttons if they are placed anywhere other than the bottom right of the screen or if they blend into the page background (NNG).

frame1.png

Dell's mobile site faced this issue when its floating chat button was positioned on the left, leading users to navigate to the ‘Contact Us’ page instead. To prevent this, Ask DaxP’s chat button is designed with a high-contrast color scheme and positioned on the right side of the application, ensuring immediate visibility.

frame2.png
Transparency: Letting Users Know They Are Chatting with a Bot

To reduce unnecessary back-and-forth exchanges, Ask DaxP provides smart suggestions through buttons and quick replies, where a study done by Luke Wroblewski on AskLukeW Chatbot stated that

Ask LukeW starts off with a series of suggested questions that change each time someone loads the page. This not only helps with the "what should I ask?" problem of empty text fields but is also a compelling way to explore what the site has to offer. Of course, someone can start with their own specific question. But in testing, many folks gravitate to the suggestions first, which helps expose people to more of the breadth and depth of available content.

When they chat with a human, some users do try to be polite to get a better response. “I try to be friendly. The nicer I am, the better they will respond.” Although not everybody feels the pressure to be polite, some people will be excessively so and waste time and bandwidth with pleasantries when they think that a real person is at the other end of the line. For example, they will say “Thank you” and “Goodbye” at the end of a conversation. Or phrase their questions with “is it possible…” Many said they would feel bad to just close the chat window at the end of a session, “out of respect for the other person.”

When users know that their chat partner is a bot, they often streamline their language to make it more easily understandable. Take advantage of this natural tendency and let people know if your agent is a bot.

frame3.png
Enhancing Navigation with Smart Suggestions

To reduce unnecessary back-and-forth exchanges, Ask DaxP provides smart suggestions through buttons and quick replies, where a study done by Luke Wroblewski on AskLukeW Chatbot stated that

Ask LukeW starts off with a series of suggested questions that change each time someone loads the page. This not only helps with the "what should I ask?" problem of empty text fields but is also a compelling way to explore what the site has to offer. Of course, someone can start with their own specific question. But in testing, many folks gravitate to the suggestions first, which helps expose people to more of the breadth and depth of available content.

a strategy recommended by chatbot UX experts (Enterprise Bot). These predefined responses help users find answers swiftly, enhancing the overall chat experience.

It is extremely helpful to provide suggestions that help the user choose responses. Add the most common responses as context-independent smart buttons or context-dependent quick replies so the user won’t have to sit through a whole conversation to get the answer they’re looking for but instead, can navigate there with a few, quick clicks.

Streamlining Conversations: No Small Talk

Politeness is essential, but efficiency is paramount in chat interactions. Studies show that when chatbots open conversations with generic phrases like “How are you today?” it can slow down interactions, leading users to delete and retype their questions (NNG). Instead, Ask DaxP adopts a direct approach with greetings like “Hello! How can I assist you today?” to keep interactions focused and user-friendly.

While the chat agent should be polite, an emphasis should be put on efficiency. Polite questions such as “How are you today?” or just the word “Hello” delay some people. In the study done by Luke Worblewski, he stated that,

One of our users had already started typing her question when the agent asked, “Hello, how are you?” She deleted what she had already typed and entered “Fine” instead and retyped her question.

frame4.png
Message Differentiation for Better Readability

Clarity in chatbot conversations significantly impacts user experience. A best practice observed in Dell’s chatbot implementation was the use of different colors to distinguish between user messages, bot responses, and system-generated notifications. Ask DaxP follows this principle by using distinctive color coding to separate messages, ensuring effortless readability for users.

Several users commented positively when the messages of the agent were shown in a different color than the messages of the user. Dell took this differentiation one step further by using yet a different color for automatically generated messages (labelled as coming from the System). One user said:  “I like the layout on the screen with different colors — my reply, and their reply, it’s really easy to see.”

Design System
Design Decisions.png

The visual identity reflects three core principles

  1. Professionalism: Aligns with the industry and builds user trust.

  2. Simplicity: Makes navigation and interaction easy and fast.

  3. Clean Aesthetic: Enhances focus and prevents distraction from the core task.

Provide suggested questions

The visual identity reflects three core principles

  1. Professionalism: Aligns with the industry and builds user trust.

  2. Simplicity: Makes navigation and interaction easy and fast.

  3. Clean Aesthetic: Enhances focus and prevents distraction from the core task.

frame5.png

Suggested questions played a big role in this transition. People would sometimes click on one of the suggested questions before putting in their own, but in all cases, reading suggested questions helped people understand how the site might best be used. After getting a response, one of the most important aspects for developing trust was seeing that answers had sources, which led to any documents cited in the response.

Images and Citations

Providing images with the response can help users understand better and faster

frame8.png
Conclusion

Designing a chatbot is, at its core, still a user experience challenge. What sets it apart is how we approach the interaction, focusing heavily on natural language as the primary interface, supported by thoughtful visual elements. It's less about screens and more about conversations, yet the principles of intuitive design, clarity, and empathy remain just as important.

bottom of page