The art and science of conversation design

What is conversation design?

Conversation design, also known as conversational design, is where content meets design thinking. It’s a discipline that aims to sets the scene for engaging and productive interactions between chatbots and humans. Effective conversation design creates seamless bot conversation experiences that help users gather information, self-serve on high-traffic enquiries and navigate their way to the most relevant products or information.

How conversation design helps organisations scale support and automation

Conversation design is a type of automation, and has many of the benefits that come with automate solutions – including the ability to scale.

Businesses that rely heavily on person-to-person conversations between employees and customers can level up with this scalability. Conversation design can be applied across:

• Information and support: Chatbots and virtual assistants are brilliant for handling enquiries and directing users to resources.
• Task automation: Chatbots can manage the day-to-day tasks of processing requests and transactions, make appointments or reservations, or altering profile details.
• Advice and recommendations: With advances in AI, chatbots are becoming increasingly skilled at completing nuanced conversations that represent lateral problem-solution thinking. Thus helping users discover information or products that either support the user’s drive to action or that even sit outside of it as a form of personalisation.

In the right setting, a chatbot can free up resources, reduce wait times and enhance customer satisfaction

What does the result of conversation design look like?

Conversation design teams collaborate to develop chatbots that engage as seamlessly as possible with users. The two ways conversation design is used is through text-based bots and voice assistants.

Text-based chatbots

Text-based chatbots are the most common bot and the easiest to derive value from when you already have a strong online presence. They’re great when used by:

• businesses with a physical product
• high-volume consumer-facing businesses, or
• service-oriented businesses.

Voice assistants

The other way we typically engage with chatbots is through voice. This is an area that is growing fast in popularity with more of us using voice assistants more regularly.

Some of the most popular voice assistants include:

• Alexa: Developed by Amazon and used with Amazon Echo devices, it’s integrated into many other smart home devices and third-party platforms.
• Siri: Siri is Apple’s voice assistant, available on iOS devices, Macs, and HomePod.
• Google Assistant: An Android bot used on Google Home speakers, phones, and other smart devices.

There are also chatbots developed for enterprise-scale solutions. These include IBM Watson and Microsoft Cortana. Each is made to assist in generating operational efficiency, and incorporate powerful AI and data integrations.

The differences between text and voice in conversation design

The work conversation designers do to create text-based chatbots and voice assistants is very similar. The content needs to have a different tone of voice based on the platform. That’s because users interact with text and voice interfaces differently.

Consider how you might like to receive information from a chatbot. If you’re driving and using Google Maps for directions, you don’t want to be overloaded with information when you’re driving. So, short, sharp and timely directions are best. If you’re using a chatbot at home things are more chill, and you can deal with more detail.

Voice is designed to be a natural conversation, where it feels more personal. While it sounds great it’s easy for the chatbot to frustrate the user with some of its responses. That means you need content that presents simple, clear responses the first time it’s heard. Text won’t ‘connect’ with users the same way, but it does give you a little more space to present options and share more detail.

Deciding what kind of automation is best for your intended use cases is something a conversation designer can help with from the outset.

Strengths and weaknesses of automated conversations

Every solution has its strengths and weaknesses. The primary strengths of a good chatbot or voice assistant is that it can:

• be made available 24/7 (no down time!)
• eliminate wait times and get to customers faster, and
• ensure consistency and generate clear outcomes.

It’s important to note that chatbots aren’t a complete solution to every customer enquiry. When a bit of lateral thinking is required, a human agent is going to be essential.

Part of the conversation design process is recognising the limitations of a bot. You need to know when it’s better to direct a matter to a human agent. This is an ongoing process as the capabilities of AI exponentially increase to solve more nuanced and complex problems for people.

How does the conversation design process work?

You’d be relieved to know the total process isn’t that complicated when looking at it from above. Just like anything technical though, the smaller details mean everything when it comes to getting a quality result.

Here’s an example of a typical conversation design process:

Step 1: Understand the user and use case

Begin the process by identifying their target users, and that user’s needs and preferences. This involves looking at what kinds of channels your customer has access to, such as:

• the types of enquiries they are likely to have
• how a voice or text bot might be best developed to service those needs.

Step 2: Gather technical requirements

Without getting the technical requirements for the chatbot you’re going to immediately run into issues. Not having correct data means the bot won’t be anywhere near as effective, and will fail to connect with the user.

Gathering technical requirements means identifying answers to questions like:

• How much will a user know? What are their likely goals?
• What likely obstacles will the user and the chatbot need to overcome together?
• How much will our bot know about the user?
• Where will the conversation end? And where do we send enquiries that can’t be handled by a chatbot?
• What data management platform do we/will we use and how will this affect the discoverability of content?

These questions and their answers shape the ‘facts’ of the conversation and ensure we’re getting the best results possible out of this automation. Remember: the goal is to replicate a human interaction as much as possible. Any falseness in the chat will lead to frustration from the user.

Step 3: Script sample dialogue

Humans are excellent at spotting fake human attributes. Think about how you react when you see an AI imitation of a person, or hear a synthetic voice in a scam call, you don’t embrace it. That’s why it’s necessary to build a personality for your chatbot. Personality and persona are essential to building trust between the user and chatbot. automation. It’s also an important part of getting your branding right through a channel that can feel quite direct and personal to your user.

This part of the process includes developing a tone of voice, a consistent vocabulary, and even a bit of a backstory for your chatbot. A lot of this work goes ‘unseen’ by the user, but it’s all there, and felt in every exchange.

Step 4: Design flows

You can’t just have your chatbot deliver a default generic response to every query, you’ve got to make it feel natural. Natural conversation is interactive, so it’s important that an automated conversation feels that way too. This means the user has some choice in the route the conversation takes, and the way we cater for that is the use of ‘conversation flows’.

These flows are like conversational decision trees, and they’re one of the most important tools in a conversation designers toolbox. They sequence the different paths that a conversation can take and help us steer the user in the right direction.

You might hear conversation designers talk about the ‘happy flow’. This is shorthand for the paths that are most common and most productive. Edge case enquiries can usually be either turned away, or directed to a human agent.

Step 5: Testing

Without testing, your chatbot will fail to connect with a user. One of the most important types of testing in conversation design is ‘Wizard of Oz’ testing. This is where you simulate conversations between users and your bot with test dialogue to see how the conversation plays out. The test gets its name from the famous book (and movie) where the Wizard of Oz is hidden while controlling a version of himself the main characters see.

There are platforms that make testing easy for text-based chatbots. Another effective method for testing early versions of your voice assistant is by having testers sit back-to-back (or use the phone) and read draft copy out loud. This allows you to verify the bot’s responses and identify any issues in the flow of the conversation.

Step 6: Deployment and ongoing iterations

Once testing is done, you’re ready to deploy. While the bulk of the work is already handled, there’s always something to be learned through continual monitoring and testing. Tracking progress and  adjusting as required is the best way to get the most out of these powerful automation tools.

You may also find that iterative changes to your chatbot are a great way to keep the tone current. Over time, you might also find that new flows can be developed to make the most of new backend integrations and more.

How to know if you need conversation design in your organisation

As you can see, there are many benefits to effective conversation design. If your organisation deals with a high volume of customer interactions and you want to find long-term savings in customer management, chatbots can be a useful investment.  When implemented correctly chatbots can deliver improved user experiences, cost savings, easier customer marketing management and access to insights and analytics.

To learn more about the benefits of conversation design speak to our content specialists.