Facebook is one of the bigger companies with a major push for bots. And a specific kind of bot: chatbots, which are integrated right inside Messenger. Facebook’s pitch was simple. Apps are bloated and complicated to use. Plus, no one is installing new apps these days. Instead, if you integrate your service or product into a messenger, something people use every single day, they’re much more likely to use your product or service.
But it’s not working out so well for Facebook. After more than a year of the Facebook bot push, Facebook is scaling back their bots effort after a reported 70 percent failure rate.
What’s wrong? The short story is that some fundamental design flaws in both Facebook’s implementation and the current state of chatbot technology have led to unintuitive solutions. Long version? Read on.
How Facebook Messenger Bots Work
Facebook Bots usually use some kind of natural language processing framework to understand what a user is trying to say. Depending on how sophisticated the bot is, or how well it is programmed, the bot will come back with an answer for a user’s question. If they don’t, the bot can either revert back a list of things the user can ask, or directly connect the user to a human.
Facebook Messenger bots that are well designed either work in tandem with a human or are extremely limited in the types of actions they can perform. For the latter, Facebook provides the developer with an array of tools and templates. They can use elements like call-to-action buttons, carousels, lists, images, media and more to create mini-app experiences inside the chat.
Say, for example, a user wants to buy flowers for their mother. They would start the chat by providing an address and then browse a selection of arrangements using a carousel. They can use buttons to go to different categories. Once they decide on an arrangement, they enter the delivery date, pay using the integrated payment option and that’s it. In this case, there was no human intervention necessary because the use case was very simple. The user didn’t need to download or install an app. All they had to do is add a contact in Messenger and get started.
If we take an example of an e-commerce Messenger bot, a human would come in to answer any question that the user has. So the user can browse around categories and items using the built-in tools. But as soon as they type in a question, it goes to an actual human. The bot doesn’t try to be smart and answer the question themselves.
Why They Are Failing
1. Because Most Bots Still Aren’t Smart
While the term “artificial intelligence” is thrown around like candy on a six-year-old’s birthday, most of the Facebook Messenger bots are not exactly smart. They don’t have a brain of their own. Most chatbots are hand-coded and that’s their downfall. Chatbots will give you an answer, or help you do something as long as it’s programmed, specifically. If you divert a bit from the script, they lose their (non-existent) mind. In some ways, chatbots are like those phone-trees, where you call customer service and press different buttons for different outcomes.
Funnily enough, most Facebook chatbots aren’t even about chat. When you start, you’re presented with some actions. Select something and you get another set of actions. So on and so forth. Most of the time you’re pressing buttons, browsing a long list to find an action. And if you think you can just ask the chatbot to do some specific thing for you, in a natural language, well, good luck with that.
2. Facebook Bots Are Hard to Find
Facebook has a huge bot discoverability problem, for both first time and returning users. There are more than 30,000 bots but there’s no central repository. If Facebook wants to be a bot platform, there has to be a bot store.
The second problem is a fundamental flaw with Facebook’s design approach. Bots show up in the same list as your conversations list. They get lost there.
3. They Try to Be Human and Fail Spectacularly
There’s nothing more annoying than a computer trying to be human and failing miserably. You know the feeling. You ask Siri to do something for you and instead of doing it, she comes back with a snarky reply. And all of this just to “add personality.”” Sadly, Facebook chatbot developers have been taking the same route.
They’re trying to make their bots friendlier, more approachable, and more fun. Those are noble goals. However, currently the human element feels like icing on a half-baked cake.
4. They’re Not Exactly Useful
This is the big one. Users would have gotten over the discovery problem, or the complicated structure and even the snark if these things were actually useful. By that, I mean that the chatbot solves a real problem. And for that given use case, it is a far superior option compared to using an app or a website.
For a vast majority of chatbots, that’s not the case.
The Future of Chatbots
Whenever there’s a new technology on the horizon, there’s a bit of a gold rush. Companies try to integrate it into their products, sometimes without thinking things through. This experimentation is a great thing. It’s what leads to innovation in both software and hardware design. But it also means that a majority of attempts result in failures. Which is also fine.
This is still round one for chatbots. In the last year, companies and developers are starting to learn where and how bots are the most effective, in which use cases bots actually make sense and help an end user. And as long as developers keep doing that, the future of chatbots is bright.
We’re seeing that for doing simple, repetitive, personalized and quick actions, chatbots are useful. Especially voice based chatbots. They’re not exactly a replacement for apps that are complex and heavily depended on visuals.
Do you use any chatbots or Facebook Messenger bots in real life? Do you find them useful for doing any specific things? Are you looking to develop a bot? Share with us in the comments below.
Image Credits: marcelokimura/Shutterstock