Conversational Interfaces

A conversational interface can refer to both the method: voice vs text - or the type of thing you use to interact: phone vs website on laptop. This can make the definition a bit confusing. Finally, as mentioned above, interfaces are also referred to as ‘modalities’. Modality is the term most commonly used when interfaces offer combined ways of interaction: voice and text combined is called multimodal for example.

Interfaces can come in all kinds of shapes, sizes, applications and functionalities. From a website chatbot, to a Bluetooth headset with built in voice assistant. Or a fluffy toy that can speak with children, or a refrigerator that its owner can talk with when out shopping to find out what they need to purchase.

Finally, if you chat with Gemini, Claude or ChatGPT on your laptop, you too are using a conversational interface.

Each interface comes with its own design challenges. Let’s have a look at the three main types:

Let’s start by looking at some examples of common conversational interfaces:

As you can see, most people are familiar with these interfaces, and are quite comfortable using them. Let’s look at some less common examples:

Interactive tablet based Agent in the lobby of a business for self service, for example checking in or out of a hotel, or collecting a rental car.

Robot waiter: voice based robot waiter that takes your order in a restaurant and bring you your choices. Often a multi modal solution, featuring some kind of visual menu to support the voice interface.

App based central heating agent: allowing you to interact with you home’s heating system whether you are at home or not. Usually text based.

Automotive in-car Agentic AI: usually voice based, but more often multi modal. Allowing you hands-free operation of all kinds of tools and services in your car.

As mentioned above: text based conversational interfaces usually take the form of chatbots. Much maligned in the past, they are becoming more and more embraced as useful tools for customer service, communication and even as a marketing and sales tool.

Text based interaction is very intuitive, and because the chatbot is visually available other UI elements can be used to power the conversation effectively. Think beyond buttons to video, images, carousels, interactive lists and so forth.

Chatbots can be found in different places: on a corporate website, in an app, in social media channels such as Instagram or Facebook, and in SMS or Whatsapp.

In short, wherever people use text to communicate, there is a potential for creating an Agentic conversational interface. Just make sure that you follow the CDI method, add established design patterns and achieve best results in your text based Conversational Interface.

As mentioned above, audio interfaces present a sterner challenge in general than text based interfaces. However, the results can be very good. Think of a fully interactive telephone Agent, who can answer all manner of questions - including details on a person’s account for example. And, if it can’t help this agent can route to the most appropriate human service representative.

That said: in this age of GenAI there is an added challenge: GenAI agents perform best if customers ask detailed and informative questions. However, many customers have grown used to barking single word instructions at IVR’s. This leads to an insufficient prompt for the GenAI to generate a useful response. So, it’s necessary to have a fallback for the Agent to acknowledge the customer's query and ask for more detailed information before generating a response. This can sometimes frustrate the customer.

Finally, as said before: apply the design patterns, make sure your Voice assistant has a suitable and clear voice, ensure that it understands queries and focus on error handling!

In this age of GenAI the technology is helping us create more and more effective Conversational interfaces. Familiar challenges remain and it’s important not to be fooled by some of the promises that GenAI seems to make: while it’s true that it’s easier than ever to build something that looks like it’s almost ‘there’, it still takes a lot of design, testing and optimization to launch an Agent that is truly helpful and delightful for your customers or employees.

Now that we’ve looked at the different kinds of Conversational Interfaces, we’d like to share a cautionary tale: with the rise of Agentic AI there have been a number of efforts to create gadgets that offer access to a GenAI Agent without much other functionality. And, although this may become a thing in the future, so far each of these initiatives has failed.

• ‘Humane pin’: a GenAI pin that could project text on one’s hand. And could also be interacted with through voice and a limited range of gestures. Launched in 2024, failed in 2025.
• Rabbit R1: a gadget similar to a smartphone, but without smartphone capabilities. Interestingly it featured a very analog scroll wheel. Launched in 2024, failed by 2025.
• “Friend”: an Agentic social gadget, a GenAI powered ‘friend’ most resembling a geriatric alarm button. Launched in 2025, fizzled out by 2026.

Each of these examples provide a cautionary tale about designing interfaces. The most important takeaway is that each of these products offered no conceivable improvement or functionality over existing smartphones or smart watches. However, we may expect certain design elements of each of these products to make some kind of return in the future. 

CDI provides comprehensive training, live support, coaching, and consultancy services to empower businesses in developing Conversational AI solutions that adhere to best practices. Our experienced team guides you through every step of the process, ensuring your implementation is efficient, effective, and aligned with industry standards.