Why “Smart” Chatbots are Actually Pretty Dumb
“Wow!” I thought to myself, “I want to do that.”
So I read the whole thing from top to bottom.
And you know what I found?
Well, nothing that impressive. The $10,000 interaction was actually just a simple series of questions and answers. The bot would ask, “Do you want this service?” and the lead would respond, “Yes, I do.”
Of course, the case study tried to illustrate the power of chatbots, but I just kind of sat there thinking, “I’m not that impressed.”
Consider, for instance, that the number one reason people prefer live chat is because of the immediate response — something chatbots are capable of providing.
This makes even more sense when you find out that customers hate waiting more than 60 seconds for a response.
And you can’t blame the customer. When they have a problem, they want it solved as quickly as possible.
That’s human nature. You and I are the same way.
We don’t want to wait for someone to replace our broken TV. We don’t want to wait to start using our new laptop because we don’t understand it.
We want answers, and we want them right now. End of story.
So, no. I’m not saying the robots are useless. They satisfy our desire for a quick response and easy solutions.
Which is probably why 2016 was a crazy productive year for chatbots.
But here’s what I am saying: a lot of people have a lot of good stuff to say about chatbots, and a lot of what they have to say is BS.
Here are a few.
“They’re so personal! It feels like you’re talking to a real person!”
“Chatbots make the buying experience so much more fun!”
“Chatbots are making my business loads of money!”
Sometimes, those might be true. But by and large, we are giving the robots way too much credit.
To prove it, let me explain the ways in which chatbots fail… hard, and why these robot buddies of ours are still pretty stupid.
Chatbots are still artificial… like really artificial
Everyone is freaking out because artificial intelligence has finally arrived.
We’re all excited to see how it impacts our quality of life and the world we live in. Depending on which robot movies you’ve watched, you might be scared, intrigued, or excited. Or a mix of all three.
But here’s the thing: The chatbots aren’t much to be afraid of or intrigued by because they’re still pretty dumb.
Sure, they have some cool tricks up their sleeves. Like answering easy questions or loading up preset phrases into a messaging app.
Here’s how it works. If you have a chatbot that wants to sell me socks, and I ask to see stockings instead, the bot will provide its lousy default message. Basically, the equivalent of the voice-recognition-software lady saying, “I’m sorry. I didn’t get that.”
And if you’re like me or the rest of the 6 billion people on this planet, nothing quite annoys you like that preloaded response.
Here’s an example of someone experiencing this hell.
Nothing kills a potential customer’s motivation like a crappy chatbot.
They want to buy your product, and they even found your bot and messaged it to start interacting with your brand.
But here’s what I can promise you: If this is their very first interaction with your company, you just lost a perfectly qualified customer for no other reason than your chatbot is stupid.
Now, take the same stupidity from the above example, add in an unhealthy dose of humor, and you’ve got yourself an even more uncomfortable experience.
I might just walk away from this interaction with a broken phone. There’s nothing worse than trying to interact with an incompetent robot.
The problem here is that the developers of these chatbots didn’t take the necessary time to make them smarter. Many chatbots have interactions that aren’t this bad, but these are examples of what can go wrong when the people developing the chatbot care a bit too little.
If the previous examples aren’t enough to make you cringe, consider this bot that loses a sale.
Yikes. That’s ugly. Everyone wave goodbye to the money train.
How do you avoid this same demise with your own chatbots?
Well, you need a designer who cares enough about the bot to test it and make sure that they’ve chosen the best keywords to trigger certain responses.
You see, that’s how decision-tree logic works. A certain phrase or keyword triggers a response from the chatbot.
This means that if developers don’t spend enough time adding the correct word triggers to the bot’s intelligence, it will remain simple.
And no one wants to talk with a simple chatbot. Here’s a final example of a bot that never took its IQ test.
Yes. Chatbots are great. They’ve got tons of potential. And heck, they might even be the next big thing in business.
But here’s what we know for sure: if you don’t spend the time adding in the correct keyword triggers, your chatbot isn’t going to help anyone. In fact, it’s going to lose your business money.
Chatbots live in a box
Here’s the second problem. Chatbots don’t have information from outside of their corresponding platform.
For example, a Facebook Messenger chatbot will only collect information about users from those individual interactions.
It doesn’t know if each user visits your website regularly, it doesn’t know what kind of stuff they are interested in, and it definitely doesn’t know how interested the user is in your product.
That’s a problem. A big problem.
How do I know?
Well, 75% of customers claim that they prefer personalized offers over general ones.
Why’s that a problem?
Because how is a chatbot supposed to deliver personalized offers to consumers when it knows practically nothing about them?
All it knows about is the interactions that the user has already had with it. It doesn’t know anything about your email marketing campaign with the user or your private messages.
It doesn’t really know the user, and because of that, it can’t personalize its message.
Since people want offers that are personal, this poses an issue.
But users don’t just want personal recommendations. They want to be recognized by name and known by their purchasing history.
But how do you do this? How do you give your chatbot more information on each user so that they can interact with them personally and make meaningful recommendations?
People wanted to be treated like people, and that’s something the robots have always struggled with.
But Opesta solves this problem. It keeps all of the chatbot interactions automated but adds a much-desired personal touch.
And if the above stats are right, that personal touch will make you A LOT of extra cash.
However you choose to do it, you need to make your chatbot interact on a more personal level if you want to be successful. Trust me. This just won’t cut it.
Chatbots are horrible at human conversation
You know what’s worse than a chatbot that admits it’s a bot and remains super boring in its interactions?
A bot that tries to pretend it’s a human and fails desperately.
Look, personalization is important. But that doesn’t mean that your chatbot needs to pretend that it’s anything other than what it is: a robot.
Quit restricting it and just let it be itself!
But seriously, everyone knows your chatbot is an AI, not a person. So don’t try to overdo it on the personality spectrum.
Here’s a bot that’s overdoing it.
Now, you and I know that this isn’t the bot’s fault. It’s the fault of the creator. Someone created this bot to be painfully un-funny.
And nothing is worse than being un-funny.
Well, actually, there is something worse:
A chatbot that’s completely unable to understand human language and change according to what someone wants.
Why does this happen? Don’t the chatbots know better?
Unfortunately, they don’t.
Still, chatbots are largely incapable of understanding all of the intricacies of human language. They simply function on keywords and phrases to guide users through a meaningful interaction.
But far too often, that system completely fails.
Without further personalization between the bot and the user, these painfully awkward conversations are going to keep happening.
Chatbots lack empathy
Someone notices your product in an advertisement, and they love it. They do some more research on your website and even ask their friends, and they fall more deeply in love with your brand and product.
So what do they do?
Well, they notice that you have a chatbot on Facebook Messenger and they figure that’s the perfect place to make their first purchase. They dive into some interactions with your bot and the next thing they know, the bot made an inappropriate joke or created an uncomfortable situation.
Something like this.
You probably are thinking to yourself that this kind of interaction wouldn’t hurt your relationship with the customer. After all, the customer knows that it’s just a robot and they aren’t going to blame its sins on your business and distrust your company… are they?
You might be right. Or you might be wrong.
Think about this. No, they don’t care about the bot. They know that the bot isn’t trying to offend them.
But they also know that your business is behind creating that chatbot. They know that any mistake it makes in the interaction is a fault on the part of the company.
And, in their eyes, it’s not just a fault. It’s lazy.
I’ve never met a consumer who wants to buy from a lazy business.
Since a chatbot can’t experience empathy, it can’t avoid these confrontations on instinct. You need to train it to do so.
Carefully construct your chatbot to avoid these offensive interactions, or you’ll lose customers for silly reasons.
Chatbots don’t understand human error
This point represents perhaps the most frustrating shortcoming of chatbots: they don’t understand when a user makes a mistake.
Often, when you’re typing something into a phone with a horribly annoying autocorrect system (there’s another crappy robot for ya), you’ll make a mistake. And when it comes to chatbots, they don’t know that it was a mistake.
It doesn’t just happen with autocorrect, either. Whenever you’re distracted or busy, you might type the wrong thing.
You’re probably tempted to think, “But isn’t that the fault of the human and not the chatbot?”
To some degree, yes.
But that’s forgetting something.
We’re not dealing with basic software or dial-this-button-to-go-to-this-menu phone systems. We’re dealing with artificial intelligence. Shouldn’t we hold that to a higher standard of genius?
After all, the whole point of making a robot capable of thinking for itself is to make it capable of interacting in a meaningful way with human beings. But that’s only possible if it understands our linguistic stumblings.
That’s not to say that it should read our minds, but it should at least understand when we want to change them.
Sadly, chatbots don’t understand when we make an error in our speech or judgment. Which means that, for now, our relationship with them is hindered.
It’s challenging. The promise of chatbots is met with an equally difficult hurdle to jump: they’re impersonal, and they feel fake.
I mean, you’re not tricking anyone with your recorded voicemail. No one listens to that and thinks, “Oh my gosh! He recorded this just for me!”
It’s the same with your chatbot.
But here’s the good news: there’s a solution.
You just need to take the time to make your bot smarter by making their interactions less artificial, taking them out of the box they’re living in, making them better conversationalists, and helping them prepare for human error.
Most the time, a stupid chatbot is the result of a lazy developer.
Truly smart chatbots, on the other hand, are the result of heavy personalization and developers that actually care.