Conversation design: Building AI Assistants with a Human Touch

Jurgen Gravestein
August 2023
5 min

When you think about it, conversations are at the heart of everything. Metaphorically speaking, we "converse" with everything in our environment. And everything "speaks back to us" in the sense that we respond to what we hear, see, and feel constantly. We interact with the world, with other people, with nature, and more than ever… with technology.

A version of this article was originally published on the Ultimate blog.

Conversation design is the art of designing these daily interactions with technology, especially AI assistants like chatbots and voice assistants. The people taking on this modest task are called conversation designers. Conversation designers are what UX designers were during the rise of web interfaces in the ‘90s. And you might want to hire some now, because AI and conversational interfaces are shaping up to be the next big UI paradigm shift.

CDI Services operates at the forefront of this paradigm shift. We’ve been pioneers in driving business value with conversational AI and in this blog post we’ll dive deeper into the importance of the conversation designer for your organization.


What makes a conversation successful

Progress might be propelled by technology, but adoption is driven by ease of use. You can invest in the best technology available on the market, but still build a chatbot that fails to deliver value to the business and your customers. That’s because building great chatbots is hard. Not only do you need to understand what people are saying, people also need to feel understood.

You see, people are experts at having conversations. We converse from the moment we wake up to the moment we go to sleep. It comes so naturally to us that we don’t even have to think about it, we’re like fish in the water. But, as it happens, there are many unwritten rules that govern our daily exchanges, whether that’s at the coffee shop, during a Zoom meeting, or while having casual conversation with a friend. Only by understanding these rules are we able to design elegant and meaningful experiences. You have to know the rules before you can break the rules.

That’s where the conversation designer comes in. A conversation designer is someone who leverages their knowledge of language, technology, and human psychology to turn your AI assistants into a successful conversational partner. It might not be immediately obvious, but language and psychology are deeply connected. When designing chatbots and voice assistants, this connection shouldn’t be underestimated. As humans, we view the world through a uniquely human lens. It’s the reason we see faces in the sky and why we like Disney characters so much. Research has shown that when we hear a voice, any voice, we automatically assign a personality to it – even if that voice is artificial. We make assumptions about who they are and where they’re from, and based on that we make split second decisions on whether we like them or not and whether we think they’re trustworthy.

Conversation designers take control over this process. When we design our AI assistants with a clear persona in mind, we increase the chances of winning the trust of the customer. If done right, customers will happily take your chatbot’s advice, follow its instructions, or in case a mistake happens, be more forgiving towards your company. Time and time again, we see that carefully designing these experiences will lead to better outcomes for your customers and your business.


Good design pays off, literally

You might be familiar with this excellent piece from CDI Services’ Jasper Klimbie, who kickstarted his conversation design career working on none other than the Google Assistant. A common misunderstanding, he explains, is that we need to get customers from A to B as quickly as possible. When providing someone with an explanation, it’s much better to go step-by-step: a technique better known as ‘turn taking’. Taking turns is something we do intuitively as people. We go back-and-forth to allow each other some space to breathe and process, and let’s be honest, we don’t consider it to be a conversation if someone is just having one long monologue.

When designing for turn taking, it might look like the dialogue becomes longer at first glance, but you have to remember our focus is on customer ease. Ease-of-use is a proxy for being quick. Breaking down information is much more effective than giving all information at once. We even did user testing to confirm and it turned out that this step-by-step approach was rated much higher by virtually all testers. Tellingly, none of them were able to accurately report how many steps the conversation took: instead, they all reported that it was easier. Conversation designers that know this, will design dialogues that lead to higher containment, because a higher percentage of customers understand the answer that they are given, which ultimately positively affects our bottom line.


What about generative AI?

The idea that you need to invest in good design to deliver business value is especially important as LLM-powered chatbots are about to take centerstage. In the early chatbot days, there used to be quite some magical thinking with regards to the capabilities of the technology, as if it could just effortlessly understand and respond to anything people say and do. This was due to a combination of overoptimism and a lack of understanding. It’s comparable to how we currently think and talk about generative AI. You could get the impression that large language models (LLMs) are the magic wands we’ve all been waiting for. In reality, working with LLMs is really challenging and requires a lot of trial-and-error to get right.

Back then it was also the case that many projects were initiated by developers. These were smart people that were early in adapting this new technology, but as a result many chatbots ended up being ‘overengineered’. What we mean by this, is that the human element was missing. There was little or no attention for the importance of language and psychology. With the rise of LLMs, we risk a replay of the past where we again see developers inside the organization pushing for innovation. At least, we should provide them with the necessary counterbalance, teaming them up with people that have demonstrable conversation design expertise and know how you turn a promising technology into a great product that your customers will love.

The best you can do as a future-forward organization is to build a sustainable conversation design practice. This is something that CDI Services has helped organizations like HP Inc., Vodafone, Salesforce, and others do. It prepared them for a future in which AI assistants will prove to be quintessential not only in meeting higher customer demands but also in retaining one’s competitive edge.

Feel free to reach out to us if you think you can use our help as well.


Jurgen Gravestein
Conversational AI Consultant
Share on:

Related articles