Not that long ago bots were a novel but clunky technology most looked at with wariness. Today, they are commonplace as more and more businesses adopt them for lead generation, customer service, and even for accepting orders! However, not all bots are “destined” for greatness. Only those whose creators upheld the rules of chatbot best practices and avoided the most common chatbot mistakes shall succeed…
Chatbots can be effective but they are not magic. They are only as good as you make them.
So, the question is, what do you need to do to avoid an epic chatbot fail?
Here are the chatbot mistakes you should try to avoid:
I presume (since you are reading this article in great detail) it’s safe to say you aren’t just interested in creating a chatbot as a fun extracurricular project.
You very likely want to use this tech to either catch quality leads, pump conversion rates or provide a better CX.
If this is the case, you are in danger of committing another of common chatbot mistakes: Forgetting to keep your bot user-centric.
The thing is, a lot of first-time bot creators get carried away by the sheer amount of possibilities. So much so they forget what they wanted to achieve in the first place.
Perhaps all they needed was an easy way to offer quick replies for better shopping and user experience… yet somehow they ended up with an assistant that derails the process by indulging in unnecessary small talk and ends up frustrating customers.
So, just because a bot CAN do something or just because you KNOW HOW to implement it, that doesn’t mean it needs to be there!
You need to think about what’s best for the end user, not you.
So, before anything else, chatbot best practices include asking yourself:
- What is the problem that the users are struggling with?
- How can your bot help them solve that problem?
- What might get in the way of that?
Really think these things through. The ultimate aim of a chatbot isn’t to be cool, it’s to solve a specific problem. So make it as helpful and as easy to use as possible.
The third chatbot mistake on our list is closely related to the previous point.
Still, I feel the need to bring it up separately. That is because when it comes to bots, there is a lot of pressure about the importance of personality.
It’s true that even though we know to be talking to software the conversation can’t lack human elements to be believable. Don’t take me wrong, your bot doesn’t need to be human. It can be an animal, made up a creature or a robot but it needs to be something.
Not so fast though! ✋
Personality shouldn’t be absent. BUT, it can’t be an obstacle in the way of the goal the bot was created to achieve in the first place either.
Personality should drive the interaction forward not stall it. If it’s not, you’ve got work to do!
So, what are some fail-safes you can apply to reach the right balance?
- Define 4-6 most important characterizing adjectives that will make up the core personality of your bot. Focus on the qualities that are noticeable through messaging (e.g. friendly, funny, competent, compassionate)
- Pick a character that best embodies those personality traits as well as fits your brand values. Focus on relatable aspects of that character such as lifestyle choices or a way of speaking to which your audience can relate. (Avoid specific things such as age, it’s not a critically defining feature).
- Create or find a visual representation of this being. It makes it easier to remember its personality. This representation can very well end up being the bot’s avatar.
- Ensure that the personality is present through language (accent, dialect, spelling, etc) and tone.
- If you want to include sequences that solely perpetuate the character and its personality deviating from any practical goal – give users the option to OPT IN or OPT OUT of this sequence.
Take, for instance, Milo. This little bot works for web designer named Stephen and he is a bot with a memorable attitude. Before you know it, you can be caught up in a lengthy but entertaining discussion about his past & present.
The great thing about this bot is that it gives you the option to go straight to what you came for (gated content or “About” info) or engage in a bit of entertaining “blah-blah”. Most importantly, his personality comes through even in the shortest, most straightforward version of the conversation!
3. Not Introducing Your Bot & Its Capabilities
Can you imagine thinking that you are talking to a human only to realize that it’s a bot? Even worse, can you imagine repeatedly asking for things the bot is simply not designed to do?
What a frustration!
The introduction makes up a crucial part of all conversational interfaces.
- Firstly, starting a conversation without introducing yourself is as weird online as it is in the real world.
- Secondly, it’s a nice conversational way of providing disclosure. The being on the other end of the interaction is not human – make it clear!
- Thirdly, it gives you a great chance to state what the bot can and can’t do. It’s a great way to manage user expectations and prevent negative experiences.
This might sound like a lot, but you can introduce your bot and provide guidance to the user in just a few sentences, it doesn’t have to be more than that.
For example, take a look at Slack’s bot, which immediately makes it clear that it’s a bot and explains how to use it.
Yurbi bot too, says hello and makes it very clear what kind of questions he can answer:
Since AI has become such a buzzword and phenomenon in the pop-culture, it’s important to manage user expectations. We are a LONG way from creating a bot from the movie HER. It’s best to let the user know what your bot can and can’t do so that they wouldn’t expect it to be something from a science fiction movie.
Just don’t leave the user hanging if your bot can’t help them. Make sure to provide guidance as to what they should do next. Usually, you’ll need to prompt them to reach out to a human that can help them solve their problem. We’ll discuss the importance of that in the last section of this article.
Remember the illusion of sentience?
The way your bot interacts with people should mimic the way people interact with each other via text messages. That’s what they expect when they use this particular communication medium. Don’t throw the users off by copy-pasting your usual copy or FAQ text or giving rigid, obviously canned responses.
You need to use a conversational tone. Though, at the same time, you need to ensure that it is appropriate for your target audience.
For example, the tone that works for a skateboard shop probably won’t work for a fintech startup, but it should be conversational in both cases.
Emojis are also an integral part of communicating via messaging apps. We all use them when chatting with friends or family to the point that it has become second nature to us. Your bot should incorporate them as well unless your brand image is extremely formal (it probably isn’t).
You might also want to consider taking advantage of GIFs since they are a popular way to express oneself on messaging apps, social media, and blogs, plus they add an element of humor to the interaction.
For example, here at Landbot.io, we use both emojis and GIFs, and it’s working great for us because it fits with our brand image.
You need to think about your target audience when designing a chatbot conversation. Are they immersed in the Internet culture and appreciate GIFs, memes, and jokes? Or do they see these things as juvenile and inappropriate for a business interaction? You don’t want your chatbot to be bland and boring, but you also don’t want it to be too out there.
If you are not sure about designing conversational AI, check out our in-depth guide on the principles of conversational scriptwriting for messaging chatbots. 💬💬💬
Speaking of designing conversations… It would be great if chatbots would understand all the ways that the same answer can be phrased.
However, that isn’t the case at the moment, and you need to make an effort to reduce the number of frustrating misunderstandings by asking straightforward questions which are likely to lead to predictable answers.
This might seem obvious, but it’s actually easier said than done, because using a conversational tone without adding unnecessary complexity is much more difficult than it sounds.
For example, in “Guide to a High-Converting Bot” we explain how even trying to be polite can complicate things.
For instance, to a question like: “Would you mind typing your email?” user can respond:
Yes, of course! | Sure… It’s …. | No, I don’t mind… | firstname.lastname@example.org
While all you really wanted was the email address! If you don’t phrase your questions clearly, the answers can result in a lot of unnecessary clutter and create conversational “glitches” that lead to user frustration.
When you ask a question in a straightforward manner, you limit the number of ways the user is likely to phrase the answer, which leads to fewer “I’m sorry, I can’t understand” instances.
There are chatbots that don’t really need to provide an option to talk to a human.
For example, Casper’s Insomnobot-3000 is a bot that you can hit up when you can’t fall asleep, and that will keep you company until you do (it might sound a bit weird, but it works!).
Its users are looking for a distraction from insomnia and the bot does the job just fine so there’s no need for human interaction.
However, if you are using your chatbot for customer service, chatbot best practices include having that option, otherwise, you are just asking for trouble (and unhappy customers can cause a lot of trouble!).
Michael Mills, a senior vice president for CGS’s contact center division, advocates a hybrid approach where simple requests are handled by the bot and the more complex ones are passed on to a customer service representative:
“It gives consumers the option of using the channel of their choice. These channels should work seamlessly together allowing a human agent to pick up where a customer left off with the chatbot, for example, and access information from that previous interaction. Making it easy for a customer to transition without repeating information or starting from scratch lends to successful customer experience.”
So make sure that the customer knows how to contact a customer service representative if they find themselves in a situation where the chatbot can’t help them.
Once again, don’t leave the customer hanging, provide them with the guidance that they need.
This might seem like a weird one on the list of top chatbot mistakes.
YES, chatbot technology has evolved beyond recognition and YES no-code bot builders make it ever so easy to create a 24-hour assistant for your website.
However, creating an efficient chatbot with a fluid conversation flow is a complex task even if artificial intelligence and Natural Language Processing (NLP) are out of the equation.
Well-converting bot requires smartly structured decision trees, great copy/conversation writing, tool integrations, attractive design and that’s just to start with. Not many of us are marketers, salesmen, writers, designers, and developers all in one.
You don’t need to know it all, provided you are honest with yourself about your strengths and weaknesses and act accordingly.
Teamwork can take your bot from “average” to “great.”
For instance, although Landbot app is designed to make bot building as simple and code free as possible, on average team-created bots have shorter production time and deliver:
- 68% more chats
- 24% higher conversion rate
On the other hand, if you don’t have the luxury of having a team, be sure to take advantage of all the advice and tools available to you. For instance, Landbot app offers a plethora of conversation flow templates, pre-programmed sections with trickier integrations as well as design templates that help put you on the right track.
Sure, building a really good bot all on your own is perfectly doable, especially if you need a quick and reliable solution.
But, if all you are looking to build a more complex “beast” that will carry the pressure of being the major lead generation channel or support system, consider inviting a few more brains on board! That’s what chatbot best practices call for!
You can’t fix chatbot mistakes and improve its performance if you have no idea how it’s doing in the first place.
Unfortunately, a lot of companies don’t pay enough attention to chatbot analytics. This means that they are completely in the dark when it comes to what works and what doesn’t.
Don’t make that mistake.
Use chatbot analytics, run user tests and listen to user inputs from the very beginning. That way, you’ll be able to see how users interact with the bot, where they tend to get stuck, which lines are well-received and which ones fall flat, and more.
You probably won’t get everything right from the get-go, but if you pay attention to the data, you will be able to optimize your bot overtime to make it more engaging, helpful, and fun.
Keep in mind that it’s best to see your chatbot not as a one-off task but as a continuous project.
It’s important to understand that a poorly designed chatbot is not an asset, it’s a liability.
You can’t just throw something together and expect it to work. Becoming a great chatbot developer requires attention to detail, patience and knowing when to ask for help.
When building a chatbot, give it the thought it deserves. Take the time to build it the right way. Make an effort to avoid the most common mistakes and follow chatbot best practices.
Only then you’ll be able to reap the benefits of this amazing technology.