The AI Conversation

Perceptions of conversational AI solutions—especially virtual assistants—vary widely based on individual experiences. Some appreciate their instant, reliable, and personalised service, while others find them frustratingly ineffective and lacking in sophistication. If you fall into the latter category, you’ve likely encountered a virtual assistant that sounds robotic, frequently asks you to rephrase your queries, or traps you in a frustrating loop with no clear way out. Such experiences can leave a negative impression and erode your trust in the solution’s ability to provide real help.

Why it matters: displaying intelligence and building user trust

Personality and tone are often underestimated but are crucial in shaping a positive user experience. A virtual assistant that understands contemporary conversational nuances demonstrates intelligence. Perceived intelligence translates to competency, which in turn builds user trust.

Consider your last customer service interaction. The agent likely started by introducing themselves, responded with phrases like “sure, no problem” or “let me check that for you,” and concluded the conversation by asking if they could assist further, followed by a polite closing such as “have a nice day.” These conversational elements are key to creating a smooth, human-like interaction.

Well-thought-out personality and tone can help you successfully reach your end-goal

Personality and tone are crucial elements of AI design that should not be ignored. Not only are they essential to create a positive user experience and build user trust, but they can also be used to help drive your solution’s end goal.

Determining how proactive or reactive the AI solution should be and designing the right personality archetype can make an impact on the overall success of the product.

For instance, if the goal of your AI is to help customers navigate large volumes of content on your website, a common pain point might be user confusion or frustration when searching for specific information. Implementing an AI solution with a friendly, interactive approach that proactively helps can address this issue. By providing immediate and accurate information, the AI can significantly improve the overall customer experience.

Welcome messages are crucial for transparency, scoping, and user education

The welcome message is the customer’s first impression of the AI solution. As the initial point of interaction, it can establish competency and build user trust, even before the user begins the conversation. It is important to include the following three main points:

First, the AI should introduce itself as an AI. The user needs to be clearly aware that they are interacting with an AI, not a human agent, to ensure they aren’t deceived.

Second, the AI should provide context on what it specialises in and how it can help the user. This is essential to ensure that the user has realistic expectations on the purpose of the AI solution.

Lastly, it’s best to gently remind the user on how to ideally interact with the AI solution by providing sample questions. In the case of a virtual assistant, for example, providing sample questions using full sentences will signal to the user that they can speak to the AI in natural language, rather than chatting to it as if it were a search engine.

We encourage users to avoid providing one- or two-word inputs to the AI, as these brief responses, like those in a Google search, don’t clearly convey the user’s intent.

For example, if a user types “my order” to an AI service agent for a clothing company, the AI lacks context to determine the exact request. Is the user seeking a tracking number, trying to cancel, or asking about a refund? A well-designed AI would ask clarifying questions to understand the user’s intent. Educating users on how to interact with AI helps reduce miscommunication and ensures smoother interactions.

Creating an introduction message that contains the right ingredients sets the scene for your AI solution as an intelligent, conversationally capable product.

Making AI less artificial with conversational charm

In a traditional Natural Language Processing (NLP) solution, such as a virtual assistant, it’s best to create a set of everyday conversational key phrases that the virtual assistant is likely to come across and define appropriate responses. The goal of the responses is for the AI solution to appear human-like and friendly. To do this, ensure that the conversational designer creates a large set of variations for each type of response. For example, rather than just “No problem,” try including several options that the virtual assistant may use to respond to the user, such as “Happy to help!” “My pleasure,” or “Anytime!”.

For Generative AI solutions, prompt engineers should include instructions on conversing with users in a way that is human-like, along with providing appropriate responses to chit-chat and off-topic questions.

By incorporating these human-like touches, you enhance the user experience, making interactions with your AI solution more enjoyable and effective.

Navigating off-topics

It’s common for users to test the limits of AI with off-topic questions. Handling such interactions requires the AI to acknowledge the broader context while gently steering the conversation back to the intended purpose. This approach demonstrates intelligence without encouraging further irrelevant questions.

Cultural norms

Keeping in mind specific cultural tendencies when designing the conversational flow of an AI solution will also affect the overall user experience. It is likely that this will come naturally for those that are designing conversational flows within their home country, however, this aspect may need to be given more attention for teams that are developing a product in a country that they are not too familiar with.

User acceptance testing

User acceptance testing is a critical aspect to ensuring that the conversational design – and overall solution – resonates with the end user. In this case, it’s important to test the solution with individuals that are representative of the actual end users and target audience for the product. Depending on user feedback, you can modify the conversational design or processes as needed to create a positive user experience.