Say hi to Kindly — our new platform
This article is based on the presentation held by our design lead, Janni Frederiksen Kalafatis, on our user forum 3 December 2018. It’s translated into English by Maja Olsen.
It’s finally here. Convertelligence is excited to introduce our new platform, which has had a thorough makeover and is launched under a new brand name — Kindly. Kindly is our product, and should be differentiated from the company we work for, which is Convertelligence.
Our product, Kindly, is quick, accessible and continuously improving — just like our chatbots. Without humans, our technology falls short. Kindly is artificial intelligence, activated by humans — now with an even more user-friendly design.
Below is a list of Kindly’s many new and exciting features.
New front-end solution
We have built the whole front-end system from scratch. By “front-end”, we mean what the user sees and how it’s connected to the back-end services. One of the benefits of building it from scratch, is that it has made the platform significantly faster, which enhances the user experience. Kindly no longer needs to load the page anew for every click in the platform.
Another benefit of this new front-end solution is that we now have complete insight into the whole ecosystem of components, since we have built them ourselves. The road from an idea to a finished product is much shorter now than before, and our customers can therefore expect more frequent updates in the upcoming year.
Smarter way of building chatbots
One of our main priorities this fall, has been to improve the tool for building dialogues. What we in Kindly call Build.
To make good chatbots, it is essential that precisely this part of the platform is effective and user-friendly.
Build has undergone a massive makeover. One of the functions we now introduce, is a new way of adding samples to a dialogue. When the chatbots grow bigger, and the dialogues follow, it’s often been difficult to continue working with them. It’s also been challenging to get an overview. We have solved this by making the field where you type in the sample, filter existing samples as you write.
This enables our customers to get an overview of the dialogue faster. On top of that, we have added counters to each sample.
We have also added the option of answering the end user with pictures. This is just the first step on the road to answering our users in various ways. In the long run, we want to add video, photo gallery, smart functions and much, much more.
Another new feature in Build, is that we have split “Bot replies” and “User says” into two tabs. This is, among other things, to create space for more advanced bot replies, but also to provide a less noisy building experience for our customers when they work with a dialogue. We want fewer options in the same place. Simultaneously, we have had to introduce a preview of the answer in the dialogue, to ensure that the customers always know what the end goal is.
The concept of how samples work has also changed. Where there were previously text fields, they now function as components. This means, among other things, that you now can mark one or more samples and drag-n-drop them into another dialogue. This is very practical when you for instance wish to merge dialogues, or transfer data from a skill to a chatbot. This also makes it easier to delete several samples at once.
Machine learning in the Inbox
Our updated platform also provides the possibility of seeing the confidence score of the answer the chatbot has given. This is the first step in helping our customers better understand how the machine learning functions in practice.
If an answer has a confidence score of 100%, the chatbot is very confident that it has given the correct answer. Similarly, a score of 50% means the chatbot is not so certain.
This will make it much easier to use real data to see which dialogues need improvement.
Additionally, this data can help describe the ideal level of the chatbot’s confidence threshold, and decide how uncertain the chatbot will have to be before it sends the user to a Fallback message.
By adjusting the confidence threshold correctly, you will get a chatbot that performs in a much better way. Helping our customers understand how the artificial intelligence works, will be a major focus area in the upcoming year.
Privacy — taken seriously
Another major update is on the privacy side.
2018 has been a year affected by four big letters: GDPR. General Data Protection Regulation. Four letters that have likely caused concerns and more work for many of our customers.
But we take privacy seriously in Convertelligence.
With our new update, we enable automatic deletion. Our customers will be able to choose and adjust how long they want their data stored for, before it’s automatically deleted.
We also launch what we have called a GDPR skill. This is a template with questions and answers regarding everything the end-user might want to know about their data, and how the chatbot handles that data.
As an extension to this, we launch a privacy-skill. By simply chatting with the chatbot, the end user can delete and download their own chat log. This will also be possible in Facebook Messenger, and in every single chat platform we integrate towards.
Everyone who has a chatbot in Kindly can subscribe to both these skills, without additional cost.
Easier to integrate
In the connection between Kindly and other platforms, lie great opportunities. You can easily connect Kindly to Messenger and Slack. In addition, we have now also integrated towards the manual chat system Zendesk. In that way, the transition between chatbot and manual chat becomes seamless. Going forwards, we will create corresponding integrations for other players such as Intercom, eDialog24, and many, many more.
New opportunities will also arise for the chat bubble, as we call it. These days, Messenger is launching a bubble solution of their own. This is still in beta, but we think this might become a preferred channel for many consumers when interacting with companies. We think being one step ahead is important, both for us and our customers.
Our goal is to make it easy to build good chatbots. With this new launch, Kindly has become a better tool for achieving our goal. And this is just the beginning. The platform will become even easier to use, and our chatbots will perform even better. We hope you are as excited as we are.