At its core, the creation of chatbots or voice assistants is all about understanding the user. It is our job as conversation designers to figure out what exactly he or she is searching for and give back helpful information. However, while we continue to improve our language understanding capabilities, chatbots often still don’t quite get it right, especially with the complex processes at play in a large corporate environment. In this case, it was a large financial institution, where my colleagues and I kept tripping over how there can be multiple different processes for what is ultimately the same action. Example? A blocked debit card. I won’t bore you with the exact details but, you can block your debit card in 5 different ways, and each of these 5 ways has its own process to unblock. Why keep it simple right? How do you design good dialogue when the answer to ‘how do I unblock my debit card’ turns out to be not so straightforward?
The exact purpose or goal behind a user question is called an intent. By recognizing the intent expressed in a customer’s input, the chatbot or voice assistant can choose the correct dialog flow for responding to it. Sounds pretty simple. However, in these complex corporate environments even seemingly straightforward intents can have complicated answers that need elaborate dialogue. For example: “Where is my money” is by far the most frequently asked question to our chatbot. It seems simple question, but to be able to build a dialogue we need so much more information. What money exactly? And are you the sender or receiver of said money? There is too many options, exceptions, alternative flows and possible links to already existing dialogues. In turn, these already existing dialogues can be just as complex and elaborate. I’m starting to feel dizzy again just thinking about it…
All these intents together create an extensive network of interconnected information. And I have to admit we got lost in that network A LOT. We needed a way to unravel and untangle this web and find a structural way to bring questions and answers together in a useful dialogue. How do we keep track of all the connections between these dialogues? How can we split up long complex dialogue into more manageable pieces? How do we prioritize what dialogues to write first? And how can we stay on top of all this in a rapidly growing conversation design team?
It was just a phase
To find our solution we went back to who we are doing this for: our customers (in the chaos one would almost forget). Looking at the customer journey more closely, we realized that what happens in a chatbot is more than ‘user asks question and bot gives answer’. We established that there are four different phases a chatbot can go through when helping a user. In phase one we define the problem of the user; what is his or her problem exactly? In phase two we define which solution suits this problem best. In the third phase we actually solve the problem, and finally, in phase four we follow up with additional useful information.
At this point it all came together. The levels of ambiguity we found in different intents corresponded seamlessly with the four phases we defined. We got ourselves a whiteboard and a huge stack of post-its and categorized every single intent in the corresponding phase. Doing this creates a complete map of how all the intents within a topic are linked, what type of answer they need, how to prioritize them and how to divide up complex dialogue in smaller pieces. Right there, Intent Mapping became a thing.
The chaos unravelled
By organizing all the intents in the corresponding dialogue phase, a full map of how the dialogues should be designed appeared before us. It allows us to create content faster and more structured, and it reveals priorities and quick wins. This makes it a lot easier to scale up as a team. On top of that, the smaller dialogue pieces make the conversations more flexible and tailored to individual users. They’ll have a more personal experience and we are able to help them a lot faster.
All in all, this method has many advantages for the content team as well as the business as a whole. Intent Mapping brings us one step closer to creating true understanding between the artificial brain and the human brain.