A couple of weeks ago, Endare was invited to the Microsoft Conversation as a Platform Hackfest in Lisbon. We had been invited to attend this hackathon and expand our hands-on experience with the chatbot technologies in which Microsoft has been investing a lot lately. Since Endare strongly believes in the future of conversation as a user interface, we gratefully accepted Microsoft’s invitation and traveled to Portugal’s finest to learn all about the possibilities and best practices. Here are some things that we learned from them during our stay in sunny Lisbon.
Microsoft has been putting a lot of effort in their Bot Framework, which seems to pay off well as it has grown to become one of the most popular chatbot development solutions. It could be seen as one of the many recent successful attempts to become once again an innovative player in the tech scene – carefully carried out after being introduced during the last year’s Build conference. Simultaneously, the company has focused on becoming more and more developer friendly over the past years, placing bets on open source and taking a lead role in cloud and web technologies. It is hence a perfect fit from a strategic point of view to set up a hackfest for supporting developers in creating chatbots.
Our team of three was accompanied by two Microsoft coworkers to build a small but useful project over the course of the three and a half days at the company’s Lisbon premises. Besides their help to connect with Microsoft’s proprietary Exchange and Active Directory services, we were able to enjoy their advice and visions on how bots should work. Our new team mates provided support both during the design of the user interactions and while tackling some technical challenges.
From a functional point of view, we became aware that a chatbot should ask as few questions as possible, and some questions should not be asked more than once. The added value of a chatbot comes not solely from the shift in user input – we should avoid transforming webforms into a Q&A – but rather from the conversational character. This includes remembering previously entered parameters in order to predict the user’s wishes. A challenge exists in protecting the thin line between predictions and assumptions, and the proper way to do so depends very much on the context. For example: asking for a confirmation each time before taking action is a secure way of approaching this, yet it might affect the overall user satisfaction.
Moreover, it is a common misconception that a chatbot’s sole manner of interaction is by – guess what – chatting. Although keyboard input is undoubtedly the most important way of accessing the software’s capabilities, there are a number of other options to spice up a bot’s user experience. So-called hero cards that contain buttons, images and attachments may serve as a welcome variation to the user while having a conversation with a bot. One particular use case can be found at the start of a chat session, as such a graphical component is much more inviting than an empty text field and a mere “Hi”. A bot that attempts to fully mimic a complete human interaction will sooner or later fail to do so, so we could as well use these kind of non-human components to improve the user experience.
Taking further on this, we had the opportunity to attend a short lecture on human hand-over. We learned about when it is the right time for the bot to transfer the conversation to a human operator – particularly useful for bots that focus on customer service. We were presented to the features proviced by the Microsoft’s Bot Framework that allow for a smooth transition from bot to human.
Under the hood, we got our hands on Microsoft’s Language Understanding Intelligent Service – or LUIS in short. The artificial intelligence service proved to be stronger and more reliable than Facebook’s Wit.ai which we already had been experimenting with. It should come as no surprise that LUIS seamlessly integrates with the Bot Framework, making the set-up phase of a project shorter and more efficient. Admittedly, LUIS suffered some hiccups as we were exploring its capabilities, but since our coaches were in direct contact with the development team, we were guaranteed that they were continuously stabilising the AI framework.
We returned home with a broadened vision, an improved skillset, and – why not – a slight tan. The privilege to attend this hackathon allowed us to explore further on the innovative technology that we believe chatbots are. Eager as always to help our customers to the fullest, we now can provide some expertise in this field as well. Therefore, don’t hesitate to contact us if your want to know how we can apply our newly acquired knowledge in your project.