8 Things I Learned at ChatbotConf 2017
My new ideas + everything which is currently HOT in the Chatbot world now!
I was glad to take part at this year’s ChatbotConf in Wien, as I was able to meet with many outstanding folks from the chatbot world. As a Chatbot developer agency, Chatbotize.me we had a booth on the conference expo area, which also let us talk to other exhibitors. This conf gave us so many new ideas, which wouldn’t have happened without Innotrade, our coolest partner, that helped us to get there! So I have to say thank them!
I’ve got so many impressions that I decided to write them down and also publish them in the form of this article. In this post I’d light to highlight the 8 most remarkable takeaways from the Conf.
I think this article will be quite thick with a lot of topics concentrated in the smallest possible place. Enjoy reading it, and benefit ;)
1. How a chatbot should work today
One of the key benefits of using a chatbot instead a basic application is that you instantly get detailed feedback from the user about your functionalities. If your stack is able to take advantage quickly of the feedbacks and probably even take act on them, you’ll be able to create a product tailored just by your users.
A chatbot sometimes represent a personal assistant for the user. Imagine an assistant in an electric device store. I’m sure we all have met assistants that did not know much about the devices. It is really disappointing, because you expect help from them. The same goes on with bots. It’s better if the assistant is an expert of televisions, but don’t know anything about anything else, than knowing a little piece of everything, but not being able to help for real with TVs. Because when you go into the store, in most cases you get into there with a purpose. You’d like to buy a TV, or a laptop, and you expect help of experts in it.
2. The big buzzwords: AI, NLP, Deep Learning
I talked a lot about AI in my recent presentations and in my ealier article. AI is a very important thing, but
those who are now working with AI, seem to be unclear with what AI really is.
What we call AI today, is a basically the so-called narrow-AI, which is capable to learn one thing better than a human can ever do. But only one thing at the same time. Moreover, what we call AI today in the terms of chatbots, is basically NLP, Natural Language Processing. Deep Learning is a mathematical method for producing AI systems, black-boxes, which will be able to do that one particular thing. And doing that one thing usually means making decisions.
While NLP performs very good at particular use cases such as we use them at our chatbot agency, Chatbotize.me for FAQ and easy phrases like bookings, using only NLP in a chatbot is too hard and cannot serve today’s needs. Technology is just not there yet. Everybody’s working on better and better NLP technologies, but it’s still in progress.
What I’ve learnt on ChatbotConf, validated my thoughts:
90% of chatbots does not apply NLP
under the hood. The leftover is also mostly the usage for only simple use cases. Hence, nowadays still rule-based chatbots rule, or as I call them, functional chatbots. When it comes to business chatbots, a guided conversation is much more expedient, because it leads the user to the right place.
Just to mention:
what’s really an interesting idea: contexts should somehow be developed with crowd sourcing.
This would definitely help the performance of today’s narrow-AI systems, because the problem with AI is the time-cost of development, as I have written about this in my previous article.
3. Voice vs Text
Big question, but simplest answer among all of them I think.
Voice is preferred at home, text at crowded places.
Voice of course is a bit different, because you most likely can use NLP. You cannot use easy one-tap CUI (Conversational User Interface) elements. So of course you have to develop voice bots differently. Voice and text-based bots will be two different areas for a while.
4. Everyone develops chatbots, but no one knows why
I have come to Chatbots from Big Data, and the same is happening in chatbot area what happened there. Top managers usually don’t really understand what should they do with the new technology. However, they create a budget for it, and they’ll see what it will do.
Two things can happen: they hire staff that understands the new technology (very rare), or they hire staff, but don’t have such a luck with them.
So that’s why a lot companies have built big data silos, but don’t know what to do with the data, and the same is happening with chatbots.
The point is, if you are a chatbot developer, you really have to face this issue. They will hire you, and they still won’t have idea about how do they want to use bots, but of course they know they definitely want to do it somehow!
So you have to prepare use cases, and show them how they can utilize this technology.
5. No one want to learn how a use a new interface again
Back in time, Hollywood had created movies about the future world, like Back to the Future 2, Blade Runner and so one. Most of these movies contains this picture about the future:
But where we really would like to live in, is this:
While we were into flying cars and steampunk technology, we realized that these new technologies constantly require new education, which is not necessarily easy. Just think about your grandma and her relationship with smartphones.
So the real goal is: when we are providing new technology we should do that on the same interface each times, preferably the human language. Because why would we learn a new remote controller or a new language, if it could communicate in our ordinary language? Yes, this is the future what we really want.
Basically,
in the future we’ll hopefully meet technology less, or at least won’t feel so much that it surrounds us.
6. Mind-blowing: Bot to bot communication
So here comes the topic which was the most mind-blowing for me presented by InterBot.
The idea of gathering a bunch of bots together and see what happens!
Why is it good for?
- Bots can work together with sharing their skills
- They can also be chained together to give a more robust bot
- Bots can be developed with different language and platform, but they still could understand each other
And this above enlightens why this could be really a big thing. A regular software to software or app-to-app integration need something in common, such as a platform or an interface. It can be object oriented, or RESTful, but still, we have to put a lot effort in integrating them together.
The big advantage of the bot-to-bot integration is every bot use the same simple language. They can talk to each other, they can require things from each other, they understand each other.
We don’t have to put effort in integrating them.
This is a very big surplus superior to app-to-app integrations.
In the current architecture it cannot be decentralized, thus there has to be a search-bot which has a registry about other bots.
My engineer personality inside is exulting at the moment. The innovator part of me thinks it’s a
big opportunity to step out from linear development of AI to exponential.
Or in other words, it makes creating wide-AI out from a lot of narrow-AI systems really easy.
And when we can reach it, when bots can really communicate with each other, there will be the time, when real AI will come.
7. Marketing and Chatfuel swags
First of all, take a big breath.
Ready?
Ok.
Chatfuel owns 46% of Messenger chatbots.
This actually means about 100.000 chatbots. That’s huge from such a relatively little team, so very big congrats to them. Unbelievable!
What’s more they have quite cool insights about Marketing usage of chatbots. I think I don’t need to bullshit here, so I’m just serving the pure facts:
- People love to answer questions via Messenger bots
- If you broadcast your users sometimes, you get 10–30 times more usage to your bots
- top performers are sending 1–2 broadcasted messages daily
- transactional functionality gives a 50% more possibility that users are going to come back
- Providing constant content is very important
8. Messenger, the flagship of bots
So at Messenger they are working hard on making the platform better and better! Number of chatbots and chatbot developers are constantly increasing:
- There are 100.000 chatbot developer
- And 200.000 chatbots
They experienced that chatbots are very good for some use cases, but they are not good for everything! Don’t try to convulsively chatbotize everything!
In cases when CUI is not the best, don’t be afraid to use webviews. Webviews blend into the chatflow very fluently these days! But be careful and don’t go too far. Don’t build a whole app inside the webview.
At Chatbotize.me, we love chat extensions and webviews, and it was good to see that it’s so popular in other teams, too.
Handover protocol is also a very useful thing, and we’re glad we’re able to use it, it makes communication with customer service very smooth and well-manageable.
What’s a really good news is Messages Objective is coming soon to Facebooks Ads, so we’ll be able to optimize our advertisements onto Messenger, and to our Messenger chatbot.
One other important thought is about being spammy, and the guys at the roundtable in the end mentioned that emails are spammy today, and Messenger is not, but it can change. However, there’s a very important difference between email and Messenger. Email is a protocol and it can be used by anyone, while Messenger is controlled by one of the world’s most innovative companies. We’re quite sure that they won’t let Messenger being spammy!
Ending
So thanks if you have read all the article. I would be honored if you share your own opinions, ideas below here as a comment. I will be happy to answer you!
Credits to my chatbot agency, Chatbotize.me
Chatbotize.me is a chatbot agency with international development experiences. We have created several chatbots for all kinds of businesses since 2016. Messenger Marketing, Customer service, NLP, AI or system integrations — that’s what we’re experts in. If you’re eager to know more, don’t hesitate to drop us a mail at:
