If you've ever watched The Jetsons as a kid, you probably remember Rosie. The friendly robot that did all the chores in the house of the future. Who haven't dreamt of having one of their own?
50 years after the The Jetsons were first aired, we're way closer to that. Robots won't prepare dinners or clean our houses yet, but they're doing pretty well as our personal assistants. No need to call a restaurant to make a reservation, just hand it over to your Siri or Alexa. Same goes for getting a cab. Or responding to e-mails without having to type.
Just how cool is that?
If People Love Bots So Much, Why Not Build More of Them?
Since they were first introduced, virtual assistants have been doing pretty well.
Then Facebook joined the game and announced their own one to be integrated into Messenger. They also added support for custom chatbots.
And that's when everybody started building chatbots.
What they promised sounded great. If people enjoy hanging out with Siri, why not use chatbots for communicating with customers?
Chatbot makers promised that they will help companies get more leads, earn more money and automate the expensive customer support.
People wouldn't have to navigate through web pages to get the information they need. They could just ask, right?
Instead of filling online forms, they could just have a casual chat with an assistant. Feels way more natural. And of course, all of that without a human on the other side.
The problem is that it doesn't work.
Bots Aren't Fun to Talk with
Do you remember the last time when you had to "talk" to a call-center robot? Like when calling your bank to activate a credit card.
Press 1 to report missing credit card... Press 2 to report a fraud... (...) Press 8 to activate a credit card...
It's far from an enjoyable experience. When you call them and hear a recorded voice, you can tell that your call is going to take a while. You also know that there's a high chance you'll end up talking to a real person anyway.
Chatbots are no better than these robots.
Pretending to Be Human and Failing at It
Do you know this kind of "customer support specialist" who rigorously follows his script?
When asked a question that's outside of it, you'll either hear "I'm sorry, I cannot help you with that." or a ridiculously generic "Have you tried turning it off and on again?".
Not really helpful, is it?
You wouldn't call them if you were facing a well-known problem - the internet does a pretty good job in helping people solve these. You reach out, because nobody has a clue what to do and you expect them to use their expertise to figure out a non-standard solution.
Put in that position, bots won't do any better. They won't improvise to help a customer - in the best scenario they'll redirect such conversation to a real person. Nothing to gain here.
They Don't Adapt
A healthy conversation requires both sides to adapt to each other. You want to be understood and you want to understand what the other person is saying. You need to understand the bigger context from your previous conversations and what you know about the other person.
If any of these is missing, the conversation won't feel right. It will seem to you that the other person doesn't care about you.
Bots are terrible at adapting, at least with the current state of AI.
They work within a very limited context, often basing their understanding on your last sentence or the current position in their decision tree. That's why a lot of things that are obvious for humans will be ambiguous for bots. When talking to them, you usually need to be so precise that it just won't feel right.
If they were a real person, you wouldn't let them talk to customers, right?
They Present an User Interface, But It's Hidden
When you walk in into a restaurant, there's usually a menu card at the table. You can use it to easily figure out what you can order for dinner. Asking the waiter wouldn't be that convenient.
- Excuse me, do you have burgers in the menu?
- Yes, we do. I can get you a classic burger, cheese burger, bacon burger (...) and our special avocado burger.
- Sounds good, what's in the avocado one? (...)
Making even the simplest order would take ages.
A lot of companies that experiment with chatbots try to replicate the information and features they already offer via traditional UIs, like web and apps. They hope that hiding them behind a conversation with a human-like bot will make it feel more natural to their users.
But just like restaurant menu cards, conventional UIs give their users a huge advantage over chats - you can see them. When a person looks at a website, they can easily scan its headlines for the information they're looking for. The traditional menu at the top provides a guideline on what else you can find there.
When you're filling a signup form, you instantly know what information they'll need from you. You can decide if you're comfortable sharing them before you engage in filling it.
That's the convenience that chatbot-based UIs aren't able to provide. When using them, people have guesses what information and features they provide. That's so far away from convenient.
Messenger (followed by a few other chat apps) noticed this problem soon after opening their platform to chatbot developers. People were constantly confused and couldn't figure out what bots offer.
To fix this issue, Facebook introduced quick reply buttons that allow chatbots to give their users a hint on what they can do next. It does its job, but well - it moves away from the concept of chatting with a bot.
What Siri Does That Chatbots Don't
So, what's so great about virtual assistants like Siri or Alexa that makes people love them? After all they share the same flaws with other chatbots.
The answer is simple - they understand our voice.
That allows us to be lazy. Or to do more by multitasking.
Think about it - if you forget to set up your alarm clock but are already laying in bed, you don't have to get up. You can just ask your phone and it will do that for you.
Same thing when you're getting ready in the morning and want to check if you should take an umbrella with you. Or when you're driving a car and want to make a quick phone call.
No need to look at your phone or computer. No need to type.
Even if they fail to understand your command once in a while, it's not a big deal. You'll often use them for things that aren't that important for you. If they won't understand your question about the weather, you have plenty of time to try asking again when doing other things. If even that fails, you'll simply have to check it on your phone, just like the assistant didn't exist.
But besides that, chatbots are just too immature to let them talk to customers (or rather make customers talk to them).
There's a long way ahead of artificial intelligence to let us build chatbots that can help more than by just replicating the features already offered by websites and apps. Once it gets there chatbots will become a huge thing, but for now there aren't many real uses for them.
P.S. There's one thing in which bots perform quite well. Check out how you can Slack bots to remind people about things.
Did you like this?
We write about building software products, no fluff included.
Leave your email and we'll let you know when we publish a new post.