Write effective bot conversations
Overview
Writing great content is key to create an engaging chatbot your customers will love. This guide is designed to help your team craft Answers that deliver a first-rate customer experience and have a measurable impact on your KPIs. If you invest time in applying these best practices, we’re confident you’ll see a serious boost in your CSAT score and deflection rates.
It's helpful to be familiar with our content blocks before reading this document. See Serve content using Message blocks for more details.
Keep it succinct
Chatbots help customers find the answers they need, quickly. Most customers are on-the-go, or multitasking, and don’t want to read through unnecessary context to get their answer, so when you write your bot content, take care to keep it as simple as possible.
Stick to the necessary information
Instead of pasting FAQs directly into the bot, adapt the content so that it works in a conversational format. Keep only to the points that will help answer the customer's question. Remember, this content is for a small chatbot window. Keep your sentences short and sweet — prevent your customers from having to scroll up to read the entire Answer.
Focus your Answer to handle one issue. Stick to one or two sentences in a text bubble, and send no more than four text bubbles at a time.
Use bulleted lists, especially for instructions
Bullet points make Answers easier to digest. Put each point or step on its own line, so it’s easier on your customer’s eyes.
Use emojis as bullets to keep your content visual. Check marks ✅ and arrows ➡️ work great, but you can also use some fun ones that inject some humour into the bot if they align with your brand’s tone 😎
Use short sentences
Keep your Answers short: make every word count. This helps your customers quickly find the information they need.
Revise content and delete anything that isn’t necessary in getting your meaning across. Avoid filler words.
Use visuals to simplify a complicated message
Sometimes, a picture is worth a thousand words. Instead of telling your customer where to go on your website to accomplish a task, show them in a quick video or GIF: it’ll clarify your message and prevent any confusion with your instructional content.
To keep your bot accessible for all customers, include a brief summary of the visual. For more complex visuals, add a Quick Reply button that links to an Answer with a longer description.
Highlight a key section in your visuals with a red circle, box, or arrow.
Be clear
Easy-to-read instructions make for an excellent customer experience. With clear content, your customers don’t need to work hard to get the info they’re looking for.
Use emojis for key words or ideas
Emojis draw your customer’s eye to the important information, and make it easier to skim Answers quickly to find the content they need.
Plus they’re fun! Emojis help develop a more informal and friendly persona that your customers will feel more comfortable with. Of course, remember to use emojis that are on brand for your company.
Acknowledge that you are a bot
Throughout a customer’s conversation, it should be very clear to them that they are interacting with a chatbot. Without this clarification, your customers may think they are chatting with a human. Use language and visual cues to indicate that this is not the case, wherever appropriate.
Start in the Greeting Answer: this is the first message your customers see. It is the most effective spot to set customer expectations for the interaction they’re about to have. Use this time to explain how to interact with the bot, and what kind of questions the bot can answer.
Acknowledge and reset expectations when the bot can’t answer a question
It’s important to acknowledge when the bot doesn’t recognize a customer’s question. It creates a more empathetic experience for customers, and indicates whether or not the bot can answer their question.
Be clear with the customer that the bot may still be learning, and their conversation will help it improve! This positions your chatbot in a favourable way by emphasizing a feature (self-learning technology) instead of focusing on a flaw (trouble understanding their question).
While acknowledging any limitations, make sure to use this time to reset the customer’s expectations by showing them what the bot can do. Reiterate the bot’s scope by suggesting some common topics it can answer, and outline how best to interact with the bot. This will help get the conversation back on track, and lead to a more positive experience for your customers.
Start Answers with a signpost
Ada’s machine learning allows customers to interact with the bot by typing questions, sometimes just keywords, to get an Answer back.
Signposts are neutral statements at the beginning of an Answer that acknowledge the customer’s input, and outline the topic of the Answer. They reaffirm to the customer that they were understood. Without these guiding statements, a basic ‘Yes’ or ‘No’ reply from the bot may confuse the customer about the Answer they’ve received, especially if they didn’t phrase their input as a binary question.
Instead of "OK, I can help with that!" be more specific: "I can help you manage your billing preferences!" This way, if the Answer happens to be slightly off topic from what the customer was looking for, they will know right away and can rephrase their question.
If Answers are part of a decision tree, and don't have training because they can only be accessed by following a specific flow through quick replies, feel free to use more generic language like "Great!" or "Understood!" You will know exactly which step the customer saw previously. There is no chance that the Answer won't make sense!
Create a conversation
Developing delightful bot experiences increases the likelihood that returning customers will ask for help from the bot before reaching out to your support team. There’s no need to stick to a cold script: writing conversational content conveys your message better, and creates a more trusting relationship between your customers and the bot.
Use plain language
Customers come from all types of backgrounds, and have different familiarities with your business. In order to assist all customers, it’s important to use plain speech and to avoid long words, organization- or industry-specific terms, and other forms of jargon. Web standards suggest that content should be written at a 5th grade level.
Before letting your Answer into the wild, take the time to read it out loud, and ‘rehearse’ it in the test bot to see how the emojis, GIFs, and text work together. Also, try running internal focus groups and getting feedback from others within your company. You may be surprised what you find out!
Use the active voice
Writing in the active voice means putting the subject of your sentence first. The passive voice is often unclear and wordy because it buries the subject in long winding phrases.
Sentences written in the active voice are shorter, tighter, and easier to read. They also address your customers directly, and make the chat feel natural.
Be empathetic
Customers contacting support have a need that is not being fulfilled. When they come to the chatbot, there is a good chance that they are already frustrated. Conveying genuine empathy in your Answers makes your customers feel understood and respected.
A canned Answer is more likely to heighten their frustration and escalate their issue to an agent (e.g., ‘We apologize for any inconvenience’).
Stay consistent
It’s important to maintain a level of consistency in your bot’s content and tone of voice. If a customer feels like the bot’s personality is inconsistent, they may be less inclined to trust what it is saying, leading to a poor customer service experience.
Make your bot a brand champion
Your customers expect a consistent experience across your brand’s services and support channels. Just like a human support agent, the bot should represent your brand’s image and values.
Take a moment to think about how your company communicates with customers on your website, in commercials, etc. Your brand voice and bot voice should be similar, but there is a distinction to be made.
While your brand could be quirky and funny in commercials, your team may decide that your customer support should be straightforward and to-the-point. Ultimately, the bot is meant to be customized to serve your brand and your company’s business goals.
Talk to your marketing team about your company’s brand guidelines. Chances are they’ll have a brand guide on-hand and would be thrilled if you followed it.
Create a tone and style
Giving the bot a personality will help you craft content that is consistent in tone and style. Once you’ve established the bot’s voice, stick to it!
If there are multiple team members working on the bot, review the content to make sure it fits the bot’s personality.
Ask yourself:
-
Does the tone fit with the bot’s personality?
-
Is your word choice similar across the board?
-
Is the formatting for lists and instructions consistent?
-
Do all your Answers sound like they’re from the same person?
Prioritize grammar and spelling
Spelling and grammatical errors look unprofessional and can damage your brand’s credibility.
Check your grammar using plugins like Grammarly, and keep a bot style guide for consistency in formatting and style. For instance, is it 1GB of data or 1 G.B.? It’s important to pick a variation that your brand will use and stick to.
A style guide should be a living document that is regularly updated.
Now that you’ve mastered writing for bots, it’s time to start designing your decision trees! Take a look at Guide chatters through conversations using Answer flows and quick replies to get started.
Have any questions? Contact your Ada team—or email us at [email protected].