When it comes to chatbot dialogue, just answering questions isn’t enough. Customers expect more.

So, if you want thousands of users interacting with your chatbot, then you better step it up.

Writing dialogues for Conversational UI is a new field. It requires the skills of a ux-designer, ux-copywriter and a creative writer.

Chatbot dialogue help customers achieve their goals and make your brand come to live. There are four fundamental components to writing purpose driven dialogues for conversational UI. It’s all about

1. what to write
2. how to write it 
3. how to measure it 
4. how to improve it

Let’s go over each point to get a better understanding of what it takes to write an effective and purpose driven chatbot. One that not only get things done for you and your customers, but is also a ton of fun to engage with.

Which chatbot dialogue should you write

Don’t get too creative too soon. Before you write anything, you need to figure out what to write.

What to write for chatbot dialogue

You can start with some clever lines trying to be the funny chatbot, but your time is better spend figuring out what your customers are trying to get done when they visit your website. A chatbot is only successful if it actually delivers value to its users, and the best way to be valuable is by solving a problem in the customer experience.

Best place to start?

Web traffic and previous interactions with your customers – you can get those from customer service. Use this information to identify the most important customer journeys. These are the journeys that your chatbot is going to improve.

Take the journey and develop it into a flowchart. From that flowchart, you take the best case scenario. Now develop a dialogue for that scenario that is going to make that scenario 10 times better.

Once you know which dialogue to focus on, set yourself some targets and get to work. The next step is to write purpose driven dialogues that ensure customers achieve their goals in record speed – and in an engaging way.

How to write a chatbot dialogue

It’s not what you say, it’s how you say it. You can deliver dull lines to your customers and be helpful. Sure, but you want to do more than that.

How to write a chatbot dialogue

Create an exciting experience that builds trust and loyalty, while helping your customers. The chatbot is an extension of your brand. And it therefore needs a well defined character with its own voice.

Best place to start?

Company style guide and character development. Here’s a trick that we’ve learned from screenwriters and novelist: if you want a character to come to live, it needs to be a real person. Get your bot a personality and backstory, and then use dialogue to make it come to live.

Your tone-of-voice document is based on the general corporate style guide and the character you develop. But please make sure this is always in development. Tone-of-voice changes over time and gets better and better.

Analyse how customers interact with the bot and adjust the tone to resonate with them more. Try different things. Run ab-tests to see which sentences get a higher conversion in your goal to help your customers get jobs done.

How we measure chatbot dialogue

Congrats. You wrote helpful and engaging dialogue and you pushed it live. Pat yourself on the back. Go home early.

…Because you need to come in early the next day. 🙂

How to measure chatbot dialogue

Conversational UI is where the ux-copywriter meets the poet. The first step of figuring out what to write was analytical, the second step of how to write was for the poet in you. Now the analytical part of your brain is getting back to work.

Always track how successful your dialogues are. Use data from your chatbot (or Intelligent Assistant that’s powering it), web-analytics, live conversations, and potentially feedback from the interface itself.

Track everything. Don’t short yourself by just looking at the number of conversations and the conversion rate at the end of the session. Analyse what happens within the conversation.

Every word is a chance to inform, engage, empathise and help. Track and test everything and commit yourself to always improving the conversation.

How we improve chatbot dialogue

When you collect a lot of data, it can be overwhelming. Thousands of interactions and questions make it tough to decide what to focus on. What you’re looking for are drop-offs.

How to improve chatbot dialogue

Moments where someone gets lost in his journey. People that get lost on your website should have engaged with the chatbot. Those who wander off mid-conversation weren’t helped by the chatbot properly. And those who ask the chatbot silly questions had trouble recognising its true value.

The answer to all these issues: ab-tests. When there are so many variables, a data driven approach is always the best approach.

If there’s a drop-off in a specific conversation, run an ab-test to optimise the dialogue. A test can be text-based, but it can really be anything that’s part of the experience.

Here’s few things you can (and probably should) test

> sentences, words, emoticons 
> the length of your answers
> the use of pictures, videos, gifs 
> forms and selection buttons
> when to trigger a proactive dialogue – and how
> how long the bot pretends to be typing

Like any ux-project, test anything you can. Analyse the interaction and within drop-offs, find isolated experiments you can run to optimise the experience.

When figuring out what to write, you focused on best case scenarios within high value journeys. When deciding what to optimise first, you look for the highest drop-offs in traffic. This is where you lose the most value.

You can definitely run multiple tests at once, as long as they are isolated. And when you tons of traffic anyway, you won’t even have to run a test that long. You’ll often have your answer the next morning.

Final thoughts on writing chatbot dialogue

It’s not easy. There are thousands of interactions and an infinite number of variations. Luckily, there’s a method to this madness.

Take these 4 steps and you should be on your way to building successful chatbot dialogues.