Bots and News
I read “Despite Stumbles, News Outlets Flock to Messenger” on The Information today and it reminded me of the “manifesto” I wrote over a year ago to help organize my thoughts around an idea about combining news and bots. I got side tracked with a different idea combining payments and chat (which ended up not working) but now I’m seeing all of these major news organizations hopping on the bot bandwagon. Maybe they are just getting caught up in the hype, which is kind of ironic given that they are writing so much about it, but I thought I’d publish my year-old manifesto for posterity. I’ve modified it slightly to make it appropriate for this audience, but give it a read and see if you believe in the future I was dreaming about a year ago.
April 25, 2015
Chat as an application has been around ever since computers were connected through the internet and has continued to evolve over the decades from internet relay chat on mainframes to instant messaging on desktops to short message service on phones.
But as you look back at all the computing platforms that have existed, the smartphone feels like the perfect fit for chat. Most people have already gravitated to it as their dominant form of communication on the phone.
Many companies have taken full advantage of this and created powerful messaging platforms with massive number of users. But there is a new trend that is emerging, where chat is morphing from just a communication medium to an interface paradigm; in other words chat is the entire interface to a product. You type in what you want, as if you were chatting with a person, and it responds to your request.
It’s incredibly simple, it works surprisingly well, and it’s an interface that everybody already inherently understands, since text messaging is so ubiquitous.
The other phenomenon that is happening is that notifications are emerging as a key interface point with users.
Notifications appear on the lock screen so the user has to do nothing to view them, just look at their phone, and if smart watches like Apple Watch and Android Wear become popular they won’t even need to take out their phones, just glance at their wrists.
Notifications are becoming the primary entry point for apps, not the grid of icons on the homescreen.
Now the beautiful thing is that chat and notifications are inherently coupled together. This is totally obvious, but important to note nonetheless. If you get a chat message from a friend, you get a notification.
So if chat is the interface and notifications are the entry point, a whole new way of how we interact with our smartphones emerges and this change is just starting to happen.
What this all means is still being defined. It’s no longer a graphical design problem, but a linguistic problem. How do you best use language to interface with the product? How do you ask the question in exactly the right way so it’s clear to the user what you are asking for and in turn they are also likely to provide a clear answer? If you think about it, even the question of “What’s your name?” is not as simple as it appears.
How to actually implement a chat based interface is also still being figured out. Some use a purely manual approach with real humans responding to every question. Some are taking a pure artificial intelligence approach using natural language processing and machine learning to construct a response. Some are using a blend of both.
And just like cloud and big data, once people accept the concept and figure out how they want to use it, you need to build core infrastructure to apply those concepts at scale. The same type of infrastructure will be needed if chat becomes the interface, especially if artificial intelligence plays a big part in helping these types of chat-based services.
Simply put, there is a big opportunity that is unfolding and there will be many ways to create value in this new space.
The hard part is picking what problem to solve first. You could potentially create any product or service through a chat interface, which is the power that this approach provides, but you need to start with a problem that is common enough for users, but also simple enough to provide a good experience with the first version.
So where do you start? What problem do you solve first?
This is probably as important of a question to figure out as the opportunity itself.
One place to start is with information.
The specific piece of information you could start with is news and even more specifically news alerts.
It is different than what most are focused on in this space (concierge services, local search, personal assistant, personal shopper), but it has some key qualities that make it an attractive starting point.
It sets up a long-term relationship with the user versus a one-time transaction. If the user wants every article about Taylor Swift, it’s not just for today but they want it continuously.
The collection of news alerts is effectively a collection of interests and by knowing their interests you can do a lot of things, such as suggesting other content they may like.
A news alert is something that people accept as a notification. It’s not like an app that is pestering you to use it again and again, but it is information the user has requested to be sent to them and often it is information that they are obsessed with maybe because they are a huge fan of a TV show or because their job depends on them being current on all the industry news.
It is fairly simple to implement, fetch articles and search for keywords, so the problem shifts towards how to setup and manage the alerts and keywords through the chat interface.
You have to chat with the user to see if they like the articles you are sending them, ask them if you’re sending too many notifications or too little, maybe ask them to suggest new publications to search for articles. All the while not being too pushy or too patronizing.
You can also explore things like changing the humor level of the responses, making the responses more sarcastic or have a more serious tone, or use different regional dialects. You can experiment making it more like talking with a friend or more like a visit to the doctor.
These are all the dimensions that make the chat interface interesting and special. They are what make it a more natural experience.
In addition, the first version of what you would build will likely be solved very manually with some basic algorithms/heuristics and will be very narrow in what it can do. But all the data that is collected from all the chat sessions with users can then be used to train and design the next version, which will be a bit more sophisticated, handle a few more edge cases, and will be able to do one more task than it could before. And this process will then repeat itself, creating a successively better and better product.
So news alerts is a potential place to start, even though it might seem trivial and mundane. But if you can make the chat interface work magically for that simple task, it is a path to make it work for a wide variety of more complex tasks.
And if you are successful in making it work for those wide variety of complex tasks, it has the potential to become the main entry point to all the capabilities of the smartphone.
And that potential could make it worth the long and hard pursuit of making it a reality.
