The mobile app explosion and sophisticated interactive web apps developed over the past 5+ years have given rise to new interface paradigms not seen before in human-computer interaction. The ability to book a car with one tap or easily search for and buy virtually any consumer good in a matter of minutes has changed paradigms. However, compromises (especially in the limited real estate of mobile devices) in those GUIs constrain their utility.
The classic compromise in GUIs is the first/naive user experience simplicity vs. the power user utility. Can you make something that a new user — especially in this era of short attention spans — can see value and potential of in 30 seconds or less? At the same time, to drive daily to weekly engagement, how do you continue to deepen that value so users regularly return?
Personalization — or, more accurately, leveraging a user’s data to tailor the experience to them individually — has been the most recent paradigm. I call these “data-drive experiences” where each user gets a unique experience best suited for their interests and behavioral patterns. However, social networks are the only large scale category of apps that have unlocked this to date. Given that they do not produce any data themselves, they leverage user-generated content from entities you are interested in as the basis of the experience. Facebook, Pinterest and Instagram have unlocked this to uncanny engagement figures. Twitter struggles with this given your “following” of an entity is only a loose marker for what interests you now in the real-time feed of world events. Discovery is a major challenge.
Outside of social networks, this hyper-individualized user interface has been challenging to realize. Most GUIs present a generic default and then ask the user to offer what they are interested in hoping the GUI can be tailored. Think of how you still have to wade through arcane filters on your bank website to show all debit transactions from last month or pick topics of interest in news aggregator feeds. It helps but is still a burden on the user to learn the interface and provide the information.
Messaging bots offer a chance to at least partially solve this problem. The great thing about not having a GUI means you are (somewhat) free to support highly individualized and arbitrarily complex interactions. Think of the weather example that has become the early poster child for bad messaging bots. The power in that interface is not asking “what is the weather today”. I can do that much quicker in a variety of versatile GUIs in one click. The power is how it can satisfy my unique needs (and then automate recurring ones) for me. Being able to say “what is the chance of rain at 8am and 6pm today” is something a weather bot should be able to answer instantly that is buried 3 layers deep in a GUI. Why? Because the GUI has to be able to solve 80% of all users’ needs at the top level. That’s success. It can’t easily solve all users’ 20% use-cases where bots can. The automation piece comes when the bot learns that’s my daily commute when precipitation is most important and starts proactively delivering it to me.
Thus, the apps vs. bots argument is a false one — except if you’re Facebook and using the Messenger platform as your next attempt at being a platform layer above the OS and apps stores. Apps and text bots are complementary to one another. GUIs best deliver information, messaging bots best capture unique requests and the automation engine customizes the GUI upon next run based on the text interaction. Its roughly what Google is doing in combining search with Google Now on Android. When I type in a query for “Mets score”, the next time I load my Now GUI, the Mets score is there now and going forward until I say I’m not interested in that as a personalized experience.
This is ultimately the most powerful interaction paradigm in interfaces we know today: GUI for display, text for interaction, automation for individualization.