For the last 5 years, we’ve been living in the app economy. There is an app for everything imaginable. Apple has ~ 1.5M apps in its app store whereas Google has 1.6M apps in the Play store. An average person uses about 25 apps a month.
Looking at these stats, it is obvious that while the ecosystem is vibrant, app discovery is broken. However, I posit that apps as a universal model for all valuable services is flawed. Ergo, the means of delivering value is as much to blame as the yet unsolved distribution & discovery problem that plagues app stores.
“There’s an app for it” is the wrong answer
An app is a designed consolidation of a few use cases that the app developer believes as valuable for users. By extension, apps try to aggregate user experience. While this increases richness of an app, it can create a frictional experience for each use case. This sort of friction forces users to break context of whatever they are doing, load a different app, and find their way through a set of screens to get an answer to whatever they are looking for.
Consider a calendar app. It does many useful things. It lets you create meetings, it lets you review your agenda, it lets you see where the next meeting is & who you are meeting. These are the core use cases of a calendar app.
However, if you probe the use cases, I’m doing a few different things in different context.
- What I do most frequently is quickly check when the next meeting starts, & find out which room its in. When I do this, I’m typically in a messaging app or window. I’m usually walking from my desk towards a room.
- While less frequent, I often look at my daily agenda twice a day. Usually, I’m waiting for coffee & catching up on messages from the night before.
- When I create meetings, I’m often in a chat thread when someone decides we need to meet face to face to discuss an item & determine next steps. I do this a few times a day.
Today, for all of these experiences, I need to open a calendar app. This not only pulls me away from what I’m doing, it also forces me through go through a frictional experience of clicking, dragging, zooming a few times — not to mention load times for each of the steps.
While the experience is less worse for infrequently used apps (because I don’t use them enough), finding them when I need them is painful.
What would be ideal is
- Deliver value to me in context of what I’m doing, without having to jump across apps.
- Discover valuable new services & get stuff done instead of wasting time & mental capacity in searching for things.
This is where bots come in.
What is a bot?
A bot is a service that listens to you when you address it, and does things that you ask for. It lives in an app where you spend a lot of time, and have a lot of conversations. This could be a messaging app, or a discussion forum, email, or a groups app.
What’s a bot good for?
Now bots won’t be the right means of delivering value for all experiences we use apps for today. I believe the right way to look at bots is examine use cases on a frequency of use & nature of use matrix. This is better than examining apps because apps aggregate user experiences & examining them might never provide a clear view of what kinds of problems bots could solve
Transactional use cases are things you can straight forward answers to, or get done without any interaction. An example here is checking where your next meeting is, who you are meeting, & its agenda.
Linear use cases require a bit more interaction, but are usually depth-first use cases with ~2 choices at each step. An example here is creating a meeting, where you choose participants, book a room or create a video conference, & an agenda.
Immersive use cases require rich interaction & often have multiple traversal paths for a user. An obvious example here is browsing your newsfeed.
Here are some more use cases examined on this framework.
Some things are obvious.
- Bots atomize experiences unlike apps which consolidate them.
- Bots often augment services that apps provide, and are a new channel of engaging users especially for daily or weekly use applications.
If I were to examine these use cases based on suitability for bots, things pan out differently for different regions.
The bot sweet spot is the region of daily use cases, that are transactional in nature. In this region, bots often augment existing full featured apps but not always. E.g. Digit, a service that saves money on my behalf sends me a daily update via SMS & offers some basic ability to ask questions. It is very primitive & doesn’t need a terrible amount of sophistication.
For monthly use cases, I believe bots can answer discovery & distribution problems if a messaging platform supports NLP as well as a page-rank type algorithm for ranking relevance of bots for specific questions its users ask. In that sense, messaging platforms can be tremendously powerful — almost like the browsers were.
For daily use cases that are immersive in nature, like browsing newsfeed, bots will offer little to no value. The magic of newsfeed or a twitter feed is in serendipitous discovery & interaction — something bots will be terrible at.
For rare use cases that are immersive, like booking personal travel & building an itinerary & making several bookings under constraints of time & money, bots will need to mature significantly at NLP and also integrate with multiple services. This is the region where a bot like M comes in.
What needs to mature?
Over the next several years, we’ll see many technologies mature which will enable teams to build bots that provide & scale entirely new services as well as augment existing apps like calendars with bots.
- Natural language processing — The power of command line interaction was in articulating exactly what you wanted & getting an answer. However, the precision & inaccessibility of commands made it difficult for an average user to use command line. With maturing NLP, the barrier imposed by precision of the query goes down.
- Voice recognition — The dominant interaction paradigm of the last decade has been touch, pinch, & zoom. However, with bots, talking actually becomes a very useful interaction paradigm. I’ve been using Amazon Echo for a few days now, and we haven’t touched our phones at home for common use cases like playing music or checking weather. It is not hard to see that eventually, we will be able to order a cab or food, or send a message via Echo.
- Messaging platforms — Messaging will be an important surface for delivering value via bots. A platform that makes it easy for the world to build bots, and users find bots that can get stuff done for them will be a $100B company. This isn’t an easy problem and requires that 1 & 2 get solved, and it also requires that the messaging platform have wide distribution in order to learn rapidly.
To summarize, I encourage all product teams to examine the use cases they are solving for, & consider bots as a first-class means of delivering value to your users. It will be an important surface for discovering new services as well as delivering a world class experience to your users.
Originally published at buildingtolearn.wordpress.com on November 17, 2015.