Chatbot Do’s and Don’ts – These Are the Best and Worst Chatbot Practices

When it comes to designing chatbots, there are a few simple practices that separate helpful, high-performing bots from chatbots you’d rather see put out of their misery.

Luckily for business owners and budding chatbot developers alike, launching a quality bot isn’t hard, as long as you know what to watch out for.

Whether you’re planning a customer support chatbot for your website or building an engaging marketing bot for Messenger, being aware of the do’s and don’ts of bot planning and development will help ensure that your chatbot is effective, and that it meets customers’ expectations every time.

We’ve compiled the best and worst chatbot practices to help you make the most of your investment, and engage in chatbot success. Use these chatbot do’s and don’ts to stay on track, helping you to design a bot that will keep users coming back for more:

1. Do: Know Your Bots
2. Don’t: Begin Chatbot Development Without a Goal
3. Do: Introduce Your Chatbot and Set Expectations
4. Don’t: Leave Users Hanging
5. Do: Give Your Chatbot a Personality
6. Don’t: Make Your Chatbot Too Cold – Or Too Hot
7. Do: Use Buttons to Speed Up Conversation and Guide Customers
8. Don’t: Forget Text
9. Do: Be Upfront About Limitations
10. Don’t: Let Your Chatbot’s Limitations Dictate its Quality
11. Do: Break Up Information as Needed
12. Don’t: Send Huge Blocks of Text
13. Don’t: Assume Conversations Will Always Move Linearly (or Smoothly!)
14. Do: Prepare for Interruptions, Ambiguities and Misunderstandings
15. Don’t: Lock Users into Conversation with Your Chatbot
16. Do: Make it Easy for Your Customer to Get Out
17. Don’t: Design a Siloed Chatbot
18. Do: Build Your Chatbot Into Your Current System
19. Don’t: Assign Chatbot Development to One Department
20. Do: Have All Departments and Stakeholders Work Jointly
21. Don’t: Rush Chatbot Design
22. Do: Edit Your Chatbot Before (And After!) It Goes Live

Do: Know Your Bots

Before writing a great novel or painting a brilliant picture, writers and artists both have the same tool in their repertoire: the references that inspired them.

Before you commit to and/or begin to design a chatbot, it’s important to understand what bots can help you with – and have some real-life references in mind.

Check out our AI chatbots and automation tools that can handle 80% of customer inquiries.

Chatbots can be designed to do almost anything. Acebot is a chatbot that prompts customers to answer a satisfaction survey in a conversational format. Kyber is a team management and organizer bot that helps your team work better together on Slack. Mila is an HR bot at Overstock that helps employees manage their schedules and even “chat”-in sick when they’re feeling under the weather. And Comm100’s AI-infused Chatbot helps customers solve simple queries, reducing queues and saving companies time and money.

Source: Acebot

Chatbots can serve other purposes, too. A quick investigation using the chatbot search engine, There Is a Bot for That, reveals a vast number of marketing, engagement, and management bots that your business can reference when planning your own chatbot.

Don’t: Begin Chatbot Development Without a Goal

Not all chatbots were created with the same end-goal in mind. Don’t begin to build your bot until you are crystal clear on what you want it to do and accomplish.

If your goal for your chatbot is to save your company money and resources, consider having it assist an individual department, such as Customer Service or HR. If your goal for your chatbot is to raise awareness of your brand, consider building a useful, engaging bot that visitors can play with.

The goal of your chatbot will affect not only the function of your bot, but also where you host it. If the goal of your chatbot is to provide customer support, consider hosting it on your website and/or mobile app. If your goal is marketing, then your chatbot will probably do best on a third-party messaging app, like Facebook Messenger or Kik. If your goal is to assist B2B clients with daily business activities, consider making your bot available on Slack.

Remember that building a chatbot into your website allows for a more customized user experience, and eliminates the risk of depending on third-party platforms. Native chatbots (or bots that are built into your company’s website or mobile app) also give companies more control over user data and allow for greater user privacy. On the other hand, chatbots on messaging platforms have greater discoverability than native chatbots, and can better keep up with customers by sending them push notifications.

Keep this in mind when considering your goals for your chatbot, and what kind of bot you want to design. Without at least one set goal in mind for your chatbot, you risk developing a less-than-effective bot on an unideal platform.

Do: Introduce Your Chatbot and Set Expectations

You may know what your bot was designed to do, but do your customers?

In order to have a successful, well-received bot, your customers need to know what your chatbot can do. (This is especially the case if your bot has limited functionality.)

Have your chatbot introduce itself, and give users a rundown of its functions. For customer support chatbots on your company’s website, this may be as simple as having your chatbot ask, “Hi, my name is ____. How can I help you?” just as a real customer support representative might.

By giving customers clear instructions on how to interact with your bot, you will keep users on-track, and asking the right questions from the start.

For example, see how CNN’s social messaging chatbot does this below:

Source: CNN

This CNN Facebook chatbot helps bring users the news – it is not a customer support bot. By listing its functionality upfront, the chatbot keeps users from being disappointed and gives them ideas and instructions on how to engage. By keeping customer expectations in check from the beginning of the interaction, CNN can give users a more satisfying experience.

When introducing your chatbot, be mindful of your bot’s greeting. Your chatbot’s greeting will create your user’s first impression of the bot. This may influence whether or not users continue chatting, and/or their overall satisfaction with the bot.

Don’t: Leave Users Hanging

Once your bot has introduced itself and invited your user to act, it’s important to respond swiftly to that action.

Don’t leave your users hanging like so:

Source: ShelfJoy

Chances are, this ShelfJoy user is not coming back.

If your bot instructs a user on how to proceed but fails to follow up with them, it teaches your user that the bot is unreliable. If your chatbot says it’s online; make sure it’s online and ready to respond to customers. Don’t leave users hanging, or else they will drop your bot on the spot.

Do: Give Your Chatbot a Personality

“…Personality is the new user experience.”

That is, according to Ultan Ó Broin from Chatbots Magazine.

Personality plays a huge roll in engaging users. When a social messaging bot has an interesting personality, users want to stick around and continue chatting with the bot. When a customer support bot has a warm, inviting personality, users feel comfortable approaching it for assistance.

To help you create a well-written bot personality, answer these simple character development questions about your chatbot:

1. Who might your brand be if it were a character? Would it be an animal or a human? How about a robot? Would it be male, female, or gender-neutral?
2. How might your brand’s character speak face-to-face with your customers? Would it be lively and positive? Helpful and warm? Straightforward and concise?
3. What attitude, tone, and style, will resonate with your customers? Are your customers millennials, baby boomers, or both? How does your target demographic communicate, and how can you tap into that to make your bot relatable?
4. What is the primary function of your chatbot? Is it to help customers with their issues? (If so, try a bot that is friendly but professional.) Is it to keep customers engaged? (If so, consider a bot that is lively and fun).

For an example of an outlandishly on-brand bot personality, look no further than the following chatbot:

Source: RuPaul’s Drag Race

Although this tone, personality, and lingo most likely won’t be ideal for your brand, this popular TV show’s Facebook Chatbot mimics its host’s charismatic personality, making for a fun and engaging user experience.

Don’t: Make Your Chatbot Too Cold – Or Too Hot

One of the reasons people like engaging with chatbots is they’re light and fun. Cold or overly robotic bots come off as too serious and transactional, and don’t inspire continued use.

As human beings, we respond well to things that sound human, even if they’re not. By adding a kind, or mildly amusing air to your serious bot, you will be able to prevent them from being too robotic.

Take a look at how Kasisto KAI’s banking chatbot stays formal, yet warm enough to be engaging:

Source: Quartz

While it pays to be friendly, don’t make your chatbot too peppy, or you run the risk of making it obnoxious. You don’t want your chatbot’s personality to outshine – or ruin – the service that you are providing, as one disgruntled user accused the weather prediction chatbot, Poncho, of doing:

Source: Poncho

Do: Use Buttons to Speed Up Conversation and Guide Customers

Buttons lets users quickly choose an option instead of typing it out. This can come in handy for chatbots – after all, it’s often faster and more convenient to engage with a single click rather than several typing clicks.

Buttons also are a great way to give customers options and show available bot offerings. They keep users on track, and within the realm of what the chatbot knows.

Dashbot, a chatbot analytics platform, has found that when bots include buttons in conjunction with quick replies, users spend more time interacting and send more messages.

Check out how personal wellness chatbot, Woebot, uses buttons to guide users:

Source: Woebot

Don’t: Forget Text

Buttons may be easier for some functions, however, don’t disregard text responses.

Text can be more engaging for customers who like to mess with bots and test what they’re capable of. It can also prove faster and easier for users who just want to tell your bot what they’re looking for, rather than click through a maze of buttons.

And, there are times when buttons cannot be implemented, such as in SMS bots like Capital One’s Eno:

Source: Fast Company

Capital One’s Ken Dodelin has no regrets about their platform’s inability to include buttons. “What’s neat here for us is that we get the raw text from the user,” Dodelin says. “You can imagine there’s a lot of learnings that can be applied when you’re seeing exactly what the customer asks, and you can build your product around that.”

While bots that use text can be a bit more time consuming to develop (their decision trees will be longer!), they become more portable and can grow far more advanced than their button-using counterparts overtime.

Recommended for you: Journey Mapping for Chatbots: How to Create a Chatbot Decision Tree from Scratch

Many effective bots use a combination of text and buttons to keep customers engaged, and successfully fulfill their functions. Check out how the weather prediction chatbot, Poncho, does this below:

Source: Poncho

Poncho gives pre-selected options to its users when asking certain questions. For example, when asked what I do in my free time, the bot gave me 3 options:

Then, it let me type a response to the specific question: “What do you do?”

This keeps the conversation with the bot flowing smoothly, while still allowing the user to give personalized responses.

Do: Be Upfront About Limitations

It doesn’t matter how many things a chatbot is designed to do: every bot has its limitations.

Source: Siri

Whether your chatbot finds its limitations in conversation or in functionality, it’s just as important to tell your customers what your bot can’t do, as it is to explain what it can do. When designing your bot, take the time to plan for the unknown.

For example, see how the Facebook Chatbot Bitcoin Buddy deals with its limitations below:

Source: Bitcoin Buddy

The best way to tell users what your chatbot can’t do is by reinforcing what your bot does well. And, like Bitcoin Buddy does here, you can also redirect users to another question to keep them within your bot’s expertise. For mobile and website bots, this may mean offering users a short selection of related customer service questions, so that they may choose the option that best matches what they’re looking for.

Don’t: Let Your Chatbot’s Limitations Dictate its Quality

Often, external limitations aren’t as much of a barrier to chatbot success as they may seem.

While you certainly want your chatbot provider to offer a robust set of features, the biggest deciding factor in the quality of your chatbot isn’t what bells and whistles are available at the time of design – it’s the quality of your copy and your decision tree.

As User Experience Designer Vaibhav Verma writes, “…when it comes to chatbots, copy is the new design. This is the ultimate mantra for designing a magical chatbot.”

The better your copy, the more engaging and helpful your chatbot. This is unconditionally true, no matter what external limitations exist. The more paths you add to your chatbot’s journey map, the more powerful it will be. And, the more information and similar questions you add, the better your chatbot will be able to assist your customers.

Recommended for you: Journey Mapping for Chatbots: How to Create a Chatbot Decision Tree from Scratch

Do: Break Up Information as Needed

There are lots of ways for your chatbot to push useful and engaging material. Depending on your bot’s purpose and design, your chatbot could send images, links, GIFs and emojis to break up information, and make the conversation more colorful.

Have your chatbot break up information by mimicking the flow of regular human messaging conversations. That means having your chatbot send users multiple short messages, instead of one long one.

For customer service bots, your bot can push links to helpful content and knowledge base articles, or can even automatically redirect users to their desired page, like Alaska Air’s support chatbot, Jenn, does:

Source: Alaska Air

Don’t: Send Huge Blocks of Text

Don’t send your chatbot users large blocks of text; they are difficult to read and may frustrate, discourage, and/or overwhelm users.

Check out the difference in how these two chatbots deliver text:

Source: Pandorabots

While this Facebook bot can do many things, effectively presenting information is not one of them. For an example of a chatbot that breaks up information well, have a look at JessBot, an accountability chatbot that is also on Facebook Messenger:

Source: Hey Jess

This chatbot also conveys a lot of information – however the information is broken up and easy for the reader to digest. This incentivizes the user to continue interacting with the chatbot, and raises satisfaction levels.

Don’t: Assume Conversations Will Always Move Linearly (or Smoothly!)

Users may want to change the topic of the conversation, or ask a follow up question. They might use sentence fragments, abbreviations, misspellings and region-specific jargon while expressing themselves.

chatbots can be prone to misunderstanding these shifts, as well as the subtleties, complexities, and colloquialisms of regular speech patterns. Here is one bot that fails to understand even the simplest conversational queues:

Source: MurphyBot

Don’t forget to account for multiple word choices, phrasings, and start options when creating your bot’s journey map, or you will end up with a frustratingly inflexible chatbot.

Do: Prepare for Interruptions, Ambiguities and Misunderstandings

As digital products designer Jesús Martín says, “Chatbots need to be designed for any possible misunderstanding in every step. That means that a specific error message needs to be set just in case the misunderstanding happens. That would help us to get the user back to the scope without restarting the whole process.”

Allow space for users to change their minds, ask another question, and/or switch from a serious conversation to simply messing with your bot. Keep your journey mapping flexible; prepare your chatbot to deal with interruptions, rather than pushing users down a rigid decision tree.

Program your chatbot to deal proactively with any ambiguities and misunderstandings. If your bot detects various key words and is unsure of a response, have it clarify with your customer before offering them an answer. Related questions are a great way to suggest a possible next step and remedy misunderstandings. That way, if your chatbot’s answer wasn’t satisfactory, your users will have several other questions to choose from that might be.

Check out how Alaska Air’s website chatbot, Jenn, masterfully handles ambiguity:

Source: Alaska Air

By responding to vague questions with multiple clickable solutions – and asking users to rephrase queries that the bot does not understand – Jenn manages to keep the conversation on track, and satisfy her users.

Misunderstanding-proofing your chatbot is a tricky task, and will likely require several attempts. Be sure to look back over chatbot data and response history to constantly create better fallback responses.

Don’t: Lock Users into Conversation with Your Chatbot

How many times have you been stuck on the phone yelling at a useless machine when all you wanted to do was speak to a representative? How many times have you tried to unsubscribe from an email marketing chain, only to continue to receive emails from the sender?

Chances are, several. And chances are, you don’t want your users to suffer through either of these bad customer experiences with your chatbot.

A successful chatbot is a tool, not an obstacle. Regardless of your chatbot’s function, don’t (accidentally or intentionally) lock your users into a conversation with your chatbot.

Source: Hey Jess

Do: Make it Easy for Your Customer to Get Out

For customer service chatbots (and other bots that are meant to adopt the function of a real human being), make sure that your customers can get from your chatbot to an agent easily. If your chatbot takes over after hours, be sure to let customers know what times they will be able to speak to a representative, should they wish to do so.

For chatbots on social messaging channels like Facebook Messenger, give your customers the option (and simple instructions on how) to start over or return to the main menu. And if your customers are subscribed to your chatbot, make it easy for them to unsubscribe and stop receiving messages.

Source: Hey Jess

Don’t: Design a Siloed Chatbot

When building a bot, companies often try to recreate functionality from scratch. This can result in an unreliable chatbot that doesn’t function in sync with the company’s existing systems.

Fabricio Teixeira, design director at Work & Co explains why this is harmful to companies and customers:

Do: Build Your Chatbot Into Your Current System

When building your own chatbot, make sure that you consider and use the systems that your company already has in place. This will give your bot greater functionality, and make it more useful.

By building the bot into your current system, you can expect less work for agents and a better customer experience.

For an example, take a look below at how effectively Starbucks’ mobile chatbot is built into its current ordering system:

Source: GeekWire

Starbucks knows that there is a growing number of purchases being made from mobile devices. In an effective response to this demand, the coffee giant has built their mobile chatbot into their preexisting system, giving it the tools and information that it needs to aid both the customer and the Starbucks team with orders.

Don’t: Assign Chatbot Development to One Department

A common yet easily avoidable mistake in bot creation is to hand exclusive responsibility of chatbot development over to the IT department. This results in a bot that may be functional as far as its technology is concerned, however, it will lack the necessary expertise and knowledge that other departments would have been able to provide.

Developing a chatbot without HR may result in a bot whose personality is not aligned with company culture. Developing a bot without Customer Service may result in a bot that leaves out important commonly asked questions, and fails to handle user queries in a way that fulfills customer expectations. Developing a chatbot without the Marketing or Social Media department may result in a bot that fails to engage its audience on social channels.

Do: Have All Departments and Stakeholders Work Jointly

Consider the function of your chatbot, and run bot development by any departments that may be able to contribute to its creation. Assign each department a task that is within their expertise in order to launch a functional, well-rounded chatbot. And keep any necessary parties in the loop for future chatbot maintenance.

When designing your chatbot’s journey map, have your employees team up to create a robust decision tree. You can do this by asking teams to come up with multiple ways that a customer might ask the same question. This can be turned into a fun group activity – whichever group can come up with the most questions in 5 minutes gets a prize!)

For example, here are some of the many ways that a user might ask a weather bot about the daily forecast:

• What’s the weather like today?
• What’s the weather like right now?
• What’s today’s forecast?
• Today’s weather
• How’s the weather?
• Show me the weather for today

And so on. By collecting possibilities from various teams, you will extend the possibilities for your users.

Don’t: Rush Chatbot Design

Some chatbot software makes bots so easy to set up, you may find yourself wanting to rush through configuration to get it live already.

Before you rush through bot design, think about the negative effects that an ill-equipped chatbot can have on your user experience. For example, a rushed customer support bot might result in reduced customer satisfaction, more customers calling in for support, and a growing number of users who will do anything to avoid using your bot a second time.

Do: Edit Your Chatbot Before (And After!) It Goes Live

Test your chatbot internally before letting your customers have at it. This will help you work out any kinks before it reaches customers. By asking users within your company to test the chatbot, you will be able to catch early on where conversation flow is breaking.

Once your chatbot has gone live, review transcripts and continue editing it. Give customers the option to rate how helpful the bot was to help you learn what is working and what isn’t.

Ensure that you don’t have any loose ends, and that the branches of your decision tree work well together.

Don’t forget to test your bot on mobile devices to make sure that it is ready to meet a mobile audience.

Conclusion

Like any new endeavor, perfecting your bot takes time, patience, and a good attitude.

By following these do’s and don’ts of chatbot design, you will be able to avoid the most common chatbot fails, and spare your company some of the growing pains of chatbot development.

For more on chatbots, check out our blog post, It’s All About The $$$ – How Much Money Can Chatbots Actually Save You?