Has it ever happened to you that end-users didn’t understand a response you prepared for your bot? Or that it turned out that a message you wrote was understood in a few different ways? You’re not alone — these are common problems all bot copywriters experience at some point or another. And while there’s no way to avoid them completely, there’s a way to write better and to anticipate these misunderstandings in advance. In this article, we’ll share with you a dialogue writing process we developed to help us and our clients write in a way that others understand.
We’ll assume you’ve already completed your requirements analysis, as well as listed out all the use cases your bot will support, and we’ll dive right into our writing process.
Part 1: Working Communication Model
Whenever we start working with our clients on bot dialogues, we want them to get a bird’s eye view of human-to-bot communication and show them what it’s like in practice. That’s why we created, what we call here, a working communication model inspired by the Gordon Model (of effective communication). It’s a great starting point for an introductory workshop on dialogue writing. Now, let’s take a look at it, shall we?
Lekta Working Communication Model
Part 2: Challenges in communication
Sounds simple? Not necessarily. As it turns out, each step of this process can pose challenges that you as a copywriter should take into account. And while they’re not unsolvable, being aware of them will help you write better dialogues.
Do you already have some ideas of what challenges we have in mind? We bet you do.
Technological challenges
For starters, if we’re dealing with a voice bot, what the user says will need to be processed by a system called ASR (automatic speech recognition). Once the exact words are recognized, another system responsible for NLU (natural language understanding) will need to interpret/make sense of the message. Even though the technology is going forward, neither of these systems is 100% reliable, which can already cause friction and lead to misunderstandings in human-to-bot communication.
Is there anything else that could go wrong? Technology aside, let’s see what other potential challenges could be waiting for us.
Human-related challenges
When a user is trying to define the meaning of her message, she also can encounter a few obstacles, making it difficult to gather her thoughts:
- not having a full understanding of an issue she’s experiencing and would like to report
- being tired or distracted
- going through emotional distress
And when she’s trying to put that message into words, there are also quite a few things that could impact the clarity of the message, and thus make it more difficult for a bot to understand:
- limited vocabulary (haven’t we all been there — experiencing a technical issue but not sure how to name the problem?)
- difficulty formulating a well-structured message
- reluctance to communicate with bots
- lack of experience communicating with bots
- certain preconceptions about how bots work
Copywriting challenges
Similarly, when a user is hearing or reading your bot’s responses, she can also be tired, distracted, or simply be in a louder place and not hear the messages well.
When she’s trying to make sense of the message she heard, she can interpret it in the wrong way too.
And while mistakes that occur because a bot is not able to process and understand users’ messages can be blamed on technology, mistakes that happen due to poorly defined meaning or poorly formulated bot messages are our human mistakes.
As bot response creators, we also:
- may not always understand the business processes or use cases we’re writing about
- or have the best understanding of the technology we’re working with
- we may be tired or distracted at work
- have difficulty choosing the right vocabulary
- or formulating a clear response
All of these can lead to less than optimal bot responses. But the good news is, that unlike all the other problems, these ones, we, as copywriters, can actually do something about.
Part 3: A new better approach to dialogue writing
A good starting point before sitting down to writing is taking a step back and fully understanding what meaning we are trying to convey. And while there’s no one universal way to do this, here’s a method we ourselves rely on and recommend to our clients.
Formulate a sample message that your bot could communicate. Let’s take “Please provide your date of birth, so I can verify your identity “ as an example.
Now, write down all the possible meanings that could be derived from this message, i.e:
- I need your date of birth
- If you want to continue, you need to provide me with your date of birth
- You need to provide me with your date of birth so that I can continue
- I know your date of birth, and I need to make sure you know it too
- Before I share sensitive information you’re asking for, I need to make sure you’re the person who you claim to be
- We do have John Smith in our system. His date of birth is 12.08.1978. You claim to be John Smith. If you know this, you probably are John Smith and I can share the information you’re asking for.
Once we realize how many possible meanings one message can have, we can decide which one of them is the most appropriate in this situation. That is, which meaning we’d like the user to take away from this message. For example, do you just want to let the user know that you need his date of birth, or do you want to stress that you need it, so you can continue?
Only after we establish the meaning of the message, we can begin to think about the right wording.
The key takeaway here is to write in a way that’s easy for others to understand — so try to make it as clear as possible. Whenever you write a bot’s response, conduct an analysis like the one above, or at least ask yourself the following questions:
- Will the end-user understand this message?
- Is there any room for misunderstanding?
- Could I make this message clearer?
- Are all the words I used necessary, taking into consideration the meaning I’m aiming for?
This process will not only help you write more clearly but also avoid simple mistakes, i.e. asking a user for an address, without specifying whether you mean permanent, temporary, or billing address. These possible misunderstandings or double meanings of your messages can definitely be discovered early if you take the right steps.
Part 4: Test & Iterate
In addition to anticipating possible misunderstandings during the writing process, you’ll still need to test your bot’s responses in real life, with real users. There’s simply no better way to see what works and what doesn’t than in practice.
Good luck! You’re now ready to test our ideas and processes in practice.
Found this article helpful and would like to learn more about the art of dialogue writing? Make sure to check out a recent article we wrote on chatbot conversational best practices. In there, you’ll find the 4 rules of successful conversation as well as 15 practical tips for great conversational experiences. Happy reading!
