Intelligent UI Requires All Disciplines
From design to implementation, various disciplines brought their experience and expertise to the table to ensure intelligence was built into SIA. It soon became clear that AI technologies alone would not make the conversational UI intelligent. This is because intelligence has many facets. It is only when different perspectives interact that SIA becomes an intelligent assistant, able to support the customer in making decisions. The various perspectives involved are outlined below.
The Concept Development Perspective
Gerlof Bachmann, Project Manager, CSS Insurance
The basic idea behind SIA is: If the digital assistant can categorize me and my intentions more accurately, it can provide me with more individual support in my search for information and communication with CSS.
Three key factors that contribute to the success of an intelligent conversational user interface emerged during the design process:
- Start small and then grow. SIA does not have to have an answer to everything from the very start. It is much more important to start by clearly defining the areas in which SIA should be used and then to learn from this and develop the interface. SIA must understand the customer journey to be able to suggest the next best action. The user should never end up at a dead end but should always be taken one step further.
- Willingness to work hard. You have to take the time to delve into customer journeys and analytics data. Developing a basis and understanding context is a time-consuming process.
- SIA’s intelligence must always be provided by a human being. To a certain degree, a system can learn, recognize patterns and take actions based on this, but it needs people who are one step ahead of the user, who accumulate experience, define meaningful action and ultimately take responsibility for the system.
The Design Perspective
Florian Stürzinger, Senior User Experience Designer, Unic
Many chatbots can only transfer information rationally and come across as very technical. Even though they tend to give accurate and generally helpful responses, they do not appeal to users on an emotional level.
SIA is entirely different: SIA is full of life. SIA engages users in a dialogue, understands their concerns, supports them and gradually gets to know them better. This creates a relationship seemingly based on trust.
With its warm colors, the user interface lends SIA a pleasant and helpful appearance. It reacts to interactions and changes its mood depending on the situation, as if you were talking to a real person.
Like an invisible companion, SIA is always at the user’s side. The SIA symbol is reminiscent of a heart rate and is embedded in the page as a floating element. It is just a mouse-click away, even if the site search is open. Wherever appropriate, SIA also offers contextual help directly within the content.
The next stage in development has already begun. SIA will soon be able to speak and listen.
The Content Perspective
Nadia Meier, Senior Content Strategist, Unic
A few weeks ago, photos of St Bernards, butlers and K.I.T.T. were up on the walls of the room dedicated to our project. (Don’t know who K.I.T.T. is? Wow, you must be very young. Please google this or ask Siri.)
What does an intelligent car from the 80s have to do with a conversational user interface from 2020? The simple answer is personality. Our aim is to bring a Smart Insurance Assistant to life: SIA, CSS’ digital helper. But who exactly is SIA? The blueprint wasn’t plucked from thin air. SIA’s purpose and key task were derived from the content strategy. Like sculptors, we molded its personality from existing material. The profile of a digital being with a human voice emerged.
And what does this voice sound like? In the absence of facial expression and gestures, personality is most often conveyed through language. How SIA speaks should be based on CSS values. That sounds good, at least in theory. But what does that mean for the content? We started writing the first dialogues and testing them out at the earliest opportunity. Does SIA speak in “plain language”? Can SIA make jokes like K.I.T.T.? Or use emojis? And dialect? The questions kept on coming. Knowing that SIA did not have to be ready from the word go reassured me. It will learn to optimize its language.
The User Experience Perspective
Dave Wick, Creative Direction, Kollektiv
Context is King. (By the way, the Queen agrees – when she is having a relaxed conversation. But that is another story.)
Customer benefit equals simplification. Services are comprehensively digitalized, they are easier to use and tasks are completed more rapidly. Every company expands its services and strives for efficiency and optimization. It is a shame, though, when the customer doesn’t want to use this simplification.
The typical chatbot is all too aware of this challenge. It is therefore quickly pushed into the corner of the screen where it leads a lonely existence. Even if it could be doing something good.
In addition to functionality and aesthetics, we focus on what encourages users to make use of their benefits: relevance. And their best friend: context. If users are presented with the opportunity to enter open dialogue at the right point in time, they feel understood. And the news gets better. Should users take this opportunity, we can then provide them with relevant questions or information.
This is why we are integrating the conversational user interface into various pages on an individual basis and depending on the content. SIA, recognizable as such, either directly addresses users or the opening question is seamlessly integrated into the page structure.
Once the obstacle of using SIA for the first time has been overcome, the conversational user interface will be able to show that it’s worth its salt. And it will have to. For more on this, let’s move on to technology.
The Technology Perspective
Dr. Clemens Blumer, Software Architect, One Inside
Wherever possible, SIA should give users a response specific to them. From a technological perspective, this is a highly complex requirement. To provide a personalized experience, SIA must be able to authenticate users. As users may have already been authenticated in other systems, it is vital that SIA does not just process data collected during the dialogue but is continuously exchanging information with surrounding systems. This requires a central component that manages and controls the flow of data and decouples systems from each other, ensuring that users are provided with personalized information, which is not shared with other systems.
When the components of a conversational UI are taken individually, their complexity is manageable. However, when all the components interact, the complexity increases exponentially. There is no single solution that can meet all these needs. We decided not to completely expand the system from the start, as we want to provide the basis for meeting all requirements and then gradually integrate more complex applications. This means that we can continually adjust the direction and available functionalities.
Training the AI is one of the main tasks and one that will continue. Even during operation, monitoring the type of use will be essential in order to react to customer needs and train the AI on an ongoing basis. Experience has shown that expected user behavior can only be correctly determined once the solution is in operation. Responding to this will be a decisive factor in ensuring that the system is perceived as intelligent, as well as using the latest technology for speech recognition.