3 Crucial Steps in Designing Conversational AI

April 8, 2024

Chatting with a bot to resolve a personal issue can be incredibly frustrating. These conversations often loop endlessly or hit dead ends after a few wasted attempts at communicating. Most product owners are aware of these issues with chatbots and understand how detrimental they can be to customer relations. This realization has prompted a significant shift toward the adoption of conversational artificial intelligence (AI), which can humanize the process of engaging with customers.

Conversational AI, at its core, is the art and science of empowering machines with the ability to understand and seamlessly respond to human language. Natural language processing (NLP) makes this possible and enables computers to imitate human interactions, learn from speech and text inputs, and translate their meaning.

Champion Advertisement
Continue Reading…

By implementing conversational AI, businesses can both reduce their operational costs and increase customer engagement. Conversational AI is also scalable. However, maintaining a personalized, empathetic touch is crucial to delivering a positive user experience.

UX designers can elevate this technology by improving conversational user interfaces (CUIs) and helping users feel supported and well understood during their interactions with chatbots. In designing conversational bots at Talentica Software, I’ve found three UX design steps to be key in solving problems and enhancing the user experience.

1. Understand the User’s Intent

In a 2022 survey, 60% of respondents answered that they would wait if that guaranteed they could chat with a human representative rather than a chatbot. Why? Because identifying and understanding user intents and tasks is easier for human representatives who can use their intuition than for bots. CUIs lack accuracy, so achieving the necessary precision can take considerable time and effort.

UX designers can add value by getting the user story right. They can improve a bot’s accuracy by identifying the user’s intent, then defining its scope. Adopting a systematic approach is an effective way of doing this. First, based on the use cases, the designers can note all the tasks and get inputs on them from possible user groups. This helps them focus on what matters and identify areas where the bot can have greater impact in the future.

When designing the CUI for a recruitment platform, I took a similar approach with the help of my team, identifying people who search for candidates, schedule interviews, provide feedback on candidates and their onboarding status, as well as people who engage with a bot for casual conversations. This was essential to creating possible conversational flows for all these roles, ensuring comprehensive coverage.

2. Map the Customer Journeys to Ensure Discoverability

Discoverability directly impacts accessibility, the user experience, and engagement. A well-designed interface in which users can easily locate services, reduces frustration, enhances task completion, and encourages active exploration and engaged interaction.

Each user is unique, responds in diverse ways, and poses questions in a variety of forms. Identifying the user’s underlying intent is the crucial, initial step, and the CUI must adeptly recognize these nuances to provide appropriate responses and prevent user frustration. For instance, in a CUI that I designed, similarities in the names of cities and people highlighted the importance of the precise identification of intent to ensure accurate responses and a seamless user experience. Figures 1 and 2 show examples.

Figure 1—Excerpt of a user’s conversation with a conversational AI
Excerpt of a user's conversation with a conversational AI

Image source: Talentica Software

To tackle this issue, designers can collaborate with skilled copywriters. This partnership can help them devise appropriate responses for similar queries while maintaining consistency with the original persona.

Designers often prioritize designing the happy paths that result in positive user experiences. However, providing solutions for the unhappy paths is equally crucial because they could lead to multiple instances of friction or interactions that run in loops, as Figure 2 shows. I have encountered prompts that had little meaning or relevance, making the identification of the user’s intent challenging.

Figure 2—An example of an unhappy path
An example of an unhappy path

Image source: Talentica Software

A reliable way of avoiding such issues is to thoroughly study the probable options that users might try, thereby reducing unwanted digressions and unhelpful experiences.

Improved discoverability also hinges on the structure of conversational flows. Designers need to consider not only existing users, but also new team members. If an existing flow proves challenging, team members might need to expend more time learning it before they can effectively contribute. This is a luxury that most businesses operating under time constraints cannot afford.

A simple visual structure for all conversational flows, with groups clearly indicated, can provide a solution. This approach can help newcomers to understand the requirements for conversational flows and learn how to navigate these flows effectively in the future.

3. Craft a Great Conversational Flow

Great chat experiences generate leads and positively impact revenues. But to craft a great user experience, designers must define the conversational elements, as shown in Figure 3. This requires that they recognize user intent, understand contexts, and be aware of the variations in language that are crucial for a more natural, intuitive interaction between users and the system. In this way, the CUI can provide accurate, relevant responses.

Figure 3—Identifying conversational elements
Identifying conversational elements

Image source: Talentica Software

As shown in Figure 4, defining the bot’s persona is equally important because this is an essential element of crafting a rich user experience. This persona encompasses everything—including what the bot would and would not say. Plus, it helps developers determine the tone and style of the chatbot’s replies. This established tone and style, in turn, assists developers in evaluating each response and maintaining coherence in communications.

Figure 4—Attributes of the chatbot’s persona
Attributes of the chatbot's persona

Image source: Talentica Software

One good approach would be to create a personality card that outlines the persona’s tone and style. Developers could then always refer to the card to check whether their responses align with the established standards.

Using Templates

The flow of a conversation can sometimes encounter barriers in template-based formats—such as when users become confused about categories of requirements. The bot might glean very little information from the templates and not learn much. This can result in unending loops that frustrate the user.

However, templated options can be useful as a fallback measure. Designers should let users write queries first so the CUI can learn from their inputs and improve its knowledge. I employed this method for the recruitment CUI, resulting in a smooth chat flow.

Avoid Dry Texting

UX designers should also pay attention to potential dead ends in a conversation, which can frustrate users. Short, curt replies, which designers commonly refer to as dry texting, do not engage people effectively and could potentially create disconnects because such responses can be repetitive and boring. A simple bye to end a chat might not convey the care a user expects from an assistant. Alternatively, a more positive and engaging might could be to say something like this: “Alright, just type help if you need anything else. Thanks for connecting with me. Goodbye.”

Final Thoughts

Creating AI experiences that are not only technologically advanced but also human centric is crucial if you are to remain relevant within the ever-evolving landscape of conversational AI. Following these three UX design steps can help simplify the process and result in intuitive, engaging, and truly transformative AI assistants. 

Principal UX Designer at Talentica Software

Pune, Maharashtra, India

Chinmay HulyalkarChinmay is a National Institute of Design, Bangalore, alumnus and has worked with companies such as Yahoo, Cognizant, and Globant. Over the last decade, Chinmay has developed expertise in product strategy, creative conceptualization, and building engaging user experiences. He has worked with both large enterprises and early and growth-stage startups.  Read More

Other Articles on Artificial Intelligence Design

New on UXmatters