Have a chat with us – how to create effective conversations for chatbots

  • Lena Tovar

Chatbots are still on the rise. Yet despite all the innovative technology – a human conversation is key. How do we create an effective chatbot conversation? Well, let’s ask Liipbot.

  • Grüezi, Bonjour and hello out there! So nice to have you here 😃 I am Liipbot and I am happy to share my expertise on how to create effective conversations for chatbots with you. Ready? So here is what I have for you:
  • Chatbots intro
    Writing tips
    Pros and Cons
    Project insight
  • Let’s start with a short intro about chatbots first. And let’s keep it crispy. Here’s what you need to know: There are mainly two types of chatbots: (1) chatbots that learn and (2) chatbots that don’t.
  • (1) Chatbots that learn are based on artificial intelligence – uh, buzzword! When they get an answer, they create algorithms out of this answer. With every further question, they verify or falsify the algorithms constantly. And that’s how those machine learning-powered chatbots can participate in complex conversations.
  • (2) Chatbots that don't learn are rule-based. They have a set of questions and answers at their disposal. Thus they follow a predefined path. This plannable set-up makes it easier from a content perspective. Usually, they are the go-to solution for businesses as they are relatively simple to implement and serve a very clear purpose.
  • So far, so good?
  • Yes. Are there any other differences?
  • Oh yes, there are! Chatbots can be found built in a messenger like Whatsapp or Slack or as a standalone application. We also distinguish between graphic interfaces like chatbots and voice assistants. Well… surprisingly I prefer chatbots 😬
  • So yes, I guess that’s it in a nutshell. Shall we continue with my ingenious writing tips?
  • Take me there.
  • Chatbots intro
    Writing tips
    Pros and Cons
    Project insight
  • Straight forward, I like that. So let’s dive right in. Ready for my tips on how to write for chatbots?
  • Yes! Let’s do this.
  • Alright. So here we go:
    Ingenious tip #1: Know your users.
    Who is interacting with your chatbot? In which context? And on which device? This is crucial because you need to know who you’re addressing your content to and in which situation. Do some user research and maybe even outline a user journey – possibly with the help of your fellow UX team members.
  • Knowing your users will have an impact on the complexity of your content as well as your tone of voice. Need an example? Be trustworthy for e-banking clients vs. speaking everyday language with lots of emojis for teenage fashion shoppers. Also, the context is essential: users in a crowded train on their way to work need succinct and quick messages vs. relaxed users on a sunday morning who even love some prose like what I do here.
  • Ready for the second tip?
  • Yes, please!
  • Well, here we go:
    Ingenious tip #2: Use conversational language.
    This might seem rather obvious but is often done wrong. When writing for chatbots, think of a real conversation and follow a conversational structure: Start with a greeting, end with a closing, ask questions in between, acknowledge and comment on the answers like yes, yay, true, no way, you name it and – very important: Don’t forget to introduce the chatbot and its purpose. It’s always nice to know to whom you are talking, right? It will lead to less awkward human-machine-interaction.
  • But most of all: Keep it short. As we know from web writing in general none of us fancies reading a super long copy. And again, think of a real conversation: Balance listening and active talking. D’accord?
  • Not really. You also made a very long point right now 😃
  • Gotcha 😅 But I am sure you too will stumble upon more complex content that can’t be explained in one sentence. In that case, there is a secret key on how to make your text more compact. Ha, curious?
  • Tell me!
  • Be implicit. As in spoken human-to-human conversations, it is not necessary to explicitly name every detail as we have all kinds of codes for situations. Of course, clarity comes first but to create a dynamic flow this so-called conversational implicature is very essential. Subconsciously our brains will decode the message based on context, experience and a cooperative mindset.
  • 🤔 Can you give an example for this?
  • Of course. Here comes one by James Giangola from the Google Conversations Design Team:

  • But why the heck should I use those conversational impli-somethings?
  • You mean conversational implicatures 🤓 They make an interaction more fun as it plays with our minds. You humans usually like that.
  • If you say so 😉 What else is useful to know?
  • Well, besides implicit phrases, there are some other elements of a smooth conversation, for example, abbreviations, exclamations, informal expressions. Even though it is a written piece, chatbot conversations shouldn’t aim at formal standards for written texts. They should feel natural. Always remember it is a conversation, right?
  • That said also don’t forget to include a sense of humour. Why? It makes conversations more lively, and it’s much more fun to interact with. But keep in mind 🚨: Only if it fits your purpose and your company’s tone of voice.
  • Sounds reasonable.
  • Exactly. This also leads to another tip:
    Ingenious tip #3: Use consistent language
    A chatbot becomes implausible when the language differs during the conversation. Cheeky and with a wink at the beginning, serious and dry at the end? Come on, that’s not comprehensible. Try to create a chatbot persona and link the conversation to all the attributes of that persona. The same applies to specific terms and vocabulary. Always use the same, consistent wording – otherwise, you will confuse your counterpart unnecessarily.
  • True. What else is relevant to know?
  • Ingenious tip #4: Avoid dead ends. Technological and language wise.
    As a copywriter who writes the chatbot conversation, always consider the flow and different paths of your storytelling. The worst thing to happen in a chatbot conversation: dead ends. Make sure to set up a valuable error message or define trigger words that lead to another part of the conversation. Please don’t use the usual «Sorry for the inconvenience bla bla», because again: You are in a conversation. Therefore it should feel like one.
  • Better go for «Damn! I am really sorry. That doesn’t seem to work.» And always provide a possible solution like «May I report that to my colleagues at the office?» so that language-wise there are no dead ends, too. Alright?
  • Alright. I am slowly getting familiar with it. What else do you have for me?
  • Cool! I take that as a compliment 😃 I still have three more advice for you.
    Ingenious tip #5: Use visual content.
    Emojis, Gifs, you name it 🤪🤑🤩🤫🤐🤤😷🥳 Whatever eases the conversation and makes it more appealing to read. Especially in longer conversations, it makes sense to sprinkle some visual content to entertain your counterpart.
  • But come on, that’s not conversational. In a face-to-face conversation, there are no emojis or gifs.
  • True. But in face-to-face you have mimics and gestures. That’s what all these visual assets are about: To add another layer of expression to your chatbot conversation. But of course only if it’s appropriate for the topic. If your chatbot is about sensitive data like medical information or your bank account one would probably be very irritated by a poo emoji 💩.
  • Hm, okay, convinced. What’s next?
  • The next tip is more about how to organise yourself when writing a chatbot conversation:
    Ingenious tip #6: Use a conversational template.
    I can tell by experience: It helps so much if you don’t just write paragraph by paragraph in your usual doc file but use a diagram to visualise the conversational flow. It quickly gets confusing when you provide several options to choose from. A template keeps you on track – and helps to identify dead ends quickly.
  • I’ve never thought of this. That’s a good tip, thanks.
  • And now, get ready for the grand finale:
    Ingenious tip #7: Do user testing.
    A proof of concept always helps, right? After drafting your conversation, it’s worth to spend some time (and 💰) on user testing. Let a group of future users have a chat with the bot. Or at least some of your colleagues or friends. You will be surprised how easily and effectively you will detect shortcomings.
  • If there is no time (and 💰) for user testing, at least read the conversation out loud. Not quite close to a user testing but at least a first indicator for barriers in the conversation flow.
  • Okay, now you know all my ingenious tips. I hope I didn’t promise too much?
  • Nope. Thanks for all the tips. I want to dig even deeper into do’s and dont’s of conversational language. Can you elaborate on that?
  • Oh, I would love to! But as said before, chatbots should be very to the point. And we have two more topics coming up. But let me recommend you this Guide to Conversational Design for more details. But now - are you ready to move on?
  • Okay, let’s continue.
  • Chatbots intro
    Writing tips
    Pros and Cons
    Project insight
  • It is always important to check potentials and risks, right? There are two perspectives on the topic: the user and the business perspective. What interests you the most?
  • The user perspective, of course.
  • Good choice 👍 From a user perspective, a chatbot offers lots of advantages. First of all: It is always accessible, 24 hours, seven days a week. That’s nice, isn‘t it? Compared to its human counterpart that’s a big plus if you’re looking for information, support or guidance.
  • Ah, talking about guidance: A chatbot also navigates the users much easier to their specific goal than conventional navigation. No need of clicking through endless pages, no need for skim reading, yay! Just an effective guided search. What a relief in our busy times.
  • And – last but not least – a chatbot adds a human touch to the plain and often mundane presentation of information on the web. It’s much nicer to use a question like «where do you live?» instead of just «address» in conversational forms. This matches with the human need for socialising and interaction and users will feel much more delighted to browse, what else can we wish for, right?
  • Absolutely! But how about the business perspective?
  • It’s all about the money, I knew it. Okay, just kidding 😂 The business perspective is important, too, and of course, your investment in a chatbot needs to pay off. And it really does. I’ll tell you why.
  • If you use the chatbot for support purposes, for example in CRM or as a sales assistant, it will save you time and money. Chatbots can handle unlimited requests at a time. In contrast, a human employee can only focus on one or maybe two requests. So service costs will be reduced, and employees can focus on the more complex and tricky cases.
  • What else? It’s also another great opportunity to strengthen your brand identity and convey your tone of voice. This can be done much easier in a conversation than in a plain text – but it is also much harder to achieve 😃 Either you create a chatbot persona that personifies a certain aspect of your brand values or one that is aiming at a specific target group.
  • Not to forget: The fame and fortune! Conversational interfaces are still a hot topic. Let’s push your innovative tech reputation 💪
  • Okay, but that were only advantages, right? Didn’t you mention cons before, too?
  • Smart you 🤓 Yes, you are right. Of course, there are some downsides or risks to keep in mind. First of all, the question should be – as in every other UX-project: Does this solution add value to my purpose? If you just want to display your opening hours, there’s no need to dive into a conversation with a chatbot.
  • The same applies to your specific business case. As mentioned before it might be inappropriate in some cases to have a chat with a chatbot when a real human is needed.
  • Another shortcoming is the limited interaction palette for rule-based chatbots. Man, that can be annoying 🙄 The solution is as simple as expensive: AI-driven chatbot. Really smart chatbots are not cheap.
  • And finally, as this is a conversation about how to write for chatbots, I have to mention the potential risk that goes with writing conversations for it. If a chatbot is not well-written meaning not user-centred, it can evoke very awkward interactions. Don’t try to be funny and cool if you are not. But thank goodness, there are experts for that 😬
  • Hehe, I got your hint 😬
  • 😃 Sorry for the hidden agenda. But my colleagues are amazing! Don’t trust me? Let me give you an example, a little sneak peek, of a real cool project they did.
  • Chatbots intro
    Writing tips
    Pros and Cons
    Project insight
  • I prefer hands-on over theory so let’s sneak into a real case. The project in a nutshell: My colleagues at Liip created a conversational chatbot called Kafibot (which translates to coffee-bot) for Zurich based energy provider Energie 360°. It was an internal communication project and aimed at bringing employees with common interests together to drink coffee. Pretty nice of e360, isn’t it?
  • Wow, that’s really cool. I would love to have that in my company, too. Why did you choose a chatbot for this?
  • Well, as the whole project encourages exchange between employees, my colleagues thought of an equally interactive and conversational approach for the process.
  • Sounds reasonable. Can I have a look at it?
  • Yes, yes, yes, so excited! Here you go 🚀

  • Isn’t that funky? Did I make you even more curious?
  • Indeed. That looks pretty cool already.
  • Well nice, then stay tuned. There will be more about the project coming soon, I promise 😃
  • Now that you got an insight into a chatbot conversation project, you know about chatbots in general and the pros and cons, and you gained some good tips on how to write for a chatbot… I think I am done with my job here. Are there any questions left?
  • Uh, I guess you answered a lot. Thanks 🙏 I feel much more confident in starting my own conversational interface project now.
  • Oh wow! I am flattered 🤗 If you want to know even more, here are some excellent resources that I used for my recommendations:

    Learn how to build your own chatbot (written by my fellow colleague Thomas Ebermann)
    Logic and conversation
    Guide to Conversational Design
    Applying Built-in Hacks of Conversation to Your Voice UI
  • Awesome. So courteous and user-centred! 😃
  • Yup, that’s my purpose. Thanks a lot, fellow chatbot friend 🙏 It was great to have a conversation with you. Arrivederci 👋

The experts behind this article
Big thanks to Pedro who magically transformed plain text in this chatbot-like conversation. Thanks to Plotti, Caro, Jenny and Janina for your valuable inputs and sharp eyes regarding copy and content. This article would not have been possible without you!


Tell us what you think