In our latest eBook about Chatbots, Conversational UI and the Future of Online Interaction we’ve included an interview with Chris Messina, Developer Experience Lead at Uber. Read the whole piece below and find out what triggered the chatbot revolution, the areas where conversational UIs are superior to native apps, how companies cheaply scale up their relationship with customers using chatbots, and the type of personality a bot should have.
Free eBook Download!
1. What triggered Bots/ Conversational UI and how did we get to the point where these are even possible?
Chris Messina: From my point of view, this is a very deep question. To answer it, I need to go back to the 60s’, when computing was just starting to be used, and bring up Douglas Engelbart´s Mother of All Demos. It’s this moment in time, where you can look back and see this vision, that was set forth, that then informed the design of computing. The context that it was stemming from had to deal with supporting the military apparatus of the Cold War. That was where a lot of the money was coming from. The Air Force was the original VC in Silicon Valley. You know, Kleiner Perkins and Sequoia Capital got started in 1972 – four years after Doug’s demo. Back then, the whole business-oriented VC world really didn’t exist as it does now. That’s why it is important to understand what was funding a lot of this stuff, and what was motivating the design of the technology that was being created. In the beginning, the technology was available only for experts. Later on, through the 70’s and then mostly starting in the 80’s with IBM and personal computers, ordinary people also had the ability to bring their work home with them. That was what led to the extinction of typewriters.
In short, the technology started out in the military and then moved into the business world. In the late 80’s and early 90’s the first cell phones emerged. They were these large bricks that were really unreliable and had these huge batteries and were really hard to use. They were also probably derived from military technology.
2007 – Steve Jobs Launches the iPhone
Fast-forward to 2007 and we have Steve Jobs launching the iPhone. At that time, Windows and Microsoft were mainly directed at professionals working in an enterprise and spending large amounts of money. Therefore, when Steve Jobs came up and said that the personal computer had finally arrived and it fit in your pocket allowing you to do everything with touch on a glass screen, it opened up the accessibility of this computing apparatus and technology that had previously been targeted solely to professionals or the military.
The three components of the iPhone were a phone that worked (which was novel at the time!) a music player (i.e. the iPod), and an Internet web browser. What was really interesting, was that this was a device that you could basically take anywhere, was connected all the time, and had a battery that lasted all day. That became kind of like a shift – a cataclysmic shift – between the previous era of business-oriented and professional-style computing to, finally, a world where personal computing became possible and desirable.
I bring this up because the transition to today’s technologies has not been a smooth, linear one. We took a lot of the metaphors that came out of that previous era of computing – file folders, and things like that, which of course directly derive from the business world – and started to adapt those models to this mobile computing operating system. You had many of the same metaphors existing, so that you could transition from that world to this new, mobile, touch-based interaction paradigm, but they’re weren’t completely at home in this new context.
“Consequently, the question is: How can brands essentially shrink themselves down into the size of a personal friend or assistant or agent or let’s say bot, so that the interaction feels as natural as talking to people that you know, through this device? That is the context that we’re all existing in.”
Slowly, as the patterns in mobile software became more apparent, we started to shift to more discrete, more asynchronous, more on-the-go types of usage of software and technology. Nowadays, instead of sitting down and having this dedicated experience with a large screen, a mouse, and a keyboard and having all these different ways of interacting with the computer, you only have this one screen that you can touch. Eventually, through Siri, you can talk to it too. This is what makes it much more accessible to a lot more people.
2016 – Conversational Commerce
Fast-forward to this year of conversational commerce, and people are getting these mobile devices to communicate and connect to each other and the people that they know. Therefore, if I can talk to my friends this way, wouldn’t it be fairly natural for me to just start talking to brands in the same manner? Brands, services, companies, and so on.
Consequently, the question is: how can brands essentially shrink themselves down into the size of a personal friend or assistant or agent or let’s say, “bot”, so that the interaction feels as natural as talking to people that you know, through this device? That is the context that we’re all existing in.
The thing that I think would be problematic for brands is to kind of imagine that this shift towards messaging really is like just another way of emailing people in a way that’s harder to block, or in a way that’s more popular. To me, it’s about building a relationship. When you think about that, it forces you to change the way that you approach your customers, the way that you get to know them, and the way that you ideally respect your customers.
Social media was a way to get free or fairly inexpensive viral spread and the ability to send a message out and then have your followers spread it and so on. The difference in messaging is that it’s a very intimate space. And if you violate the intimacy of that space, then you lose that relationship forever, because people will block you. I think in the messaging context, especially in Facebook where users have one account and brands have one Facebook page, if you start berating your customers with messages through the Messenger platform, you will get blocked and you’ll get kicked off the platform and you will lose access to that billion-plus-sized pool of users.
And once Facebook opens up WhatsApp as well — which could happen this year or next year — that’s another billion people around the world that you would have access to. I think that the dynamic is going to be really carefully controlled, because Facebook sees that it is a long-term play, to disrupt email, to disrupt SMS, to disrupt all these different mechanisms that people currently use to communicate.
2. In which areas are conversational UIs based on messenger protocols superior to native apps (and where will apps remain first choice)?
Chris Messina: When you think about what makes sense to build in the messaging space, how do you convert your app to be useful in the messaging context? I don’t think that’s actually the right question. I think the right question is: what is the service that you’re building and how do you create and support the relationship that you have with your customers through messaging? What are the set of expectations that your customers bring to that channel and how can you be responsive and useful in a way that doesn’t feel spammy or abusive? You know, if you wouldn’t want to receive the messages that you’re sending to your customers, then you should probably think twice about what it is that you’re sending and why you’re sending it.
In terms of how or what should remain as full-fledged apps, I think that there are reasons for apps to persist and to be part of the overall relationship structure. There are many different ways that we can interact with our friends and family members: over chat or voice or over Snapchat or over Facetime or email or face-to-face. I think brands need to think about themselves as being available on all those different channels as well. If you’re a smaller brand, you have to figure out which channels make the most sense to you and are the most comfortable and on which you can be the most responsive. But, of course, bigger brands have more opportunity to be as accessible as your friends are, and through all these different channels. Based on the problem that you’re trying to solve, or based on the current intent of the person, being able to switch between these different channels is very important.
For example, when I’m trying to coordinate dinner with my partner, it might make more sense for me to just call her as opposed to trying to go back and forth through 3000 text messages. I think for brands, it’ll similar where maybe someone starts a conversation, let’s say for a return of something that they bought and that happens in the messaging context. It’s very easy for me to go to Messenger and find a brand and say, “Hey, I have a problem with an order.” Great, now the brand can look that information up, they can reply to me through Messenger and then they can say, “Oh, well if you’d like to find a replacement, you know, now it’s a really good time to download our app and check that out, because that’s got our full catalog.” Trying to recreate the way in which some of these native experiences have been optimized for inside the context of a message app doesn’t necessarily make sense.
“If you wouldn’t want to receive the messages that you’re sending to your customers, then you should probably think twice about what it is that you’re sending and why you’re sending it.”
It’s about figuring out those touch-points and those hand-off-points between a conversational context, where someone has a general question or doesn’t exactly know how to express what it is that they want. First is, giving someone a directed task to do in an interface or application that’s designed specifically for that purpose. In the case of Uber, you can order an Uber from within Facebook Messenger today and I think this makes sense in the case where you have relatives visiting from someplace else and who don’t use Uber a lot. Well, suddenly, it becomes very easy, especially if you’re having a conversation with the relatives in Facebook Messenger, to say “Hey, you know here’s my address. Get a ride.” Tap that address, book them a ride, the car shows up, they don’t even have to download the app. Alternatively, for someone who uses Uber on a daily basis, having that conversation to get a car is less efficient than just opening the app, placing the pin on the map, and then going through that process which our designers of course have spent a long time working on and optimizing.
There’s a spectrum and I think that’s the important thing to realize and always put this stuff in context. Understand that now we’re finally living in an era of real personal computing where I think relationships are going to be the defining characteristic of how the software gets built out and adopted.
3. Many companies see bots or messenger services as a way to cheaply scale up this closeness to customers. Can that work or is it a trap that these companies might fall into?
Chris Messina: I think you’ve really hit the nail on the head in that the real challenge is once you start to imagine and think about brands being as accessible as your friends, of course your friends don’t scale. And so, how can companies deal with all of their hundreds or thousands or hundreds of thousands of customers in a way that actually makes sense in the messaging context, when, of course, dealing with each and every one of these conversations individually is extremely difficult and costly?
I personally know this, because I have a bot — it’s called MessinaBot — and it allows me to hold office hours, essentially. When I first tried this out, I thought that I’d be able to somehow handle ten simultaneous conversations at once — but I couldn’t deal with it. There was just too much context switching and it was very, very confusing. So, I can only imagine what it would be like for a big brand with thousands of customers to sort of manage that.
This is where I think automated systems, i.e bots, combined with some machine learning and artificial intelligence become essential. It’s this spectrum of automation-to-human-based interaction that I think makes sense. There’s always going to be a set of very basic straightforward questions or issues that most customers run into and I imagine that most customer support issues fall into probably 80% or 90% of the same kind of category. Those questions which you see on a repeated basis are the ones that an automated system should be able to handle and you should have essentially your first responder being an automated system that eventually says, “Hey look, if you’re having one of those problems, let’s go through this conversational wizard, and take care of your problem and get you on your way.” A lot of people would actually appreciate that, because they don’t want to talk to a person or several people and be handed off through a bunch of support request chains and have to tell all those people the same information, etc. So, it’s fine to use automation in that case. As long as it’s very clear and upfront and it is like “Hey, I’m an automated system. How can I help you?”
At some point, when you get into that last 10% or maybe even 5% of questions that require human intervention, that handoff I think is critical. Where it says, “Okay, I can’t help you but, I’m going to bring in someone to do that”. There are many similar patterns that already exist out there on the web and in the world. That’s when bringing humans into the loop makes sense.
In essence, that’s a very straightforward easy way to think about scaling these systems. Of course, the other thing you want to be doing is monitoring all of these conversations, both between the automated systems and, of course, the human agents and increase your machine learning and your ability to gain insights and knowledge and awareness of where the issues and where the challenges are. That input can then become a feedback loop that fits into your product cycle and your product development cycle. Consequently, if people are having a hard time, let’s say, recovering their password, maybe you need to redesign your sign in flow or something similar. These things become very useful feedback channels, which is why it’s all part of the relationship that emerges over time.
4. Could we say that it mostly is about the handover between automated responses on bots and handing it over to the real customer service agent?
Chris Messina: There is an experience aspect that I think it’s really important to evaluate, as a lot of people will think about bots and automated systems as a way of lowering their customer costs. In my opinion, that’s actually an insufficient way of evaluating whether or not this is going to succeed. For example, with Uber, we currently do not use automated bots on Facebook Messenger. If you send a message to Uber on Facebook Messenger, you will talk to a live person and the reason is we’ve got people out in the world who are looking for a car, trying to find their driver, and in those moments an automated system is only going to increase their frustration. So, you’ve got to think about how your customers are coming to this experience, what they’re bringing to the mix and whether or not they’re in a mood to kind of deal with automation or not.
5. Uber doesn’t do automated customer service through bots. Do you know of any brand you personally have experienced doing that well, both the AI part and the automated part and also the recognizing of when actually a human takes over? Is there any example that you can think of right now?
Chris Messina: It’s not really an example that you might think of, if you’re only looking at the chat space and the messaging space, but whenever you call Apple, you’re immediately put in touch with an automated system and that automated system is actually quite good, you know. I don’t know and I doubt that it’s the same technology that’s powering Siri, but it does do natural language understanding and processing and it kind of tries to help you solve whatever problem you’re having and very quickly, if you express any frustration, I’m pretty sure they’re using, a technology that sort of allows them to understand your mood or your sentiment, or whatever is in your voice. They will pass you off to a human very, very quickly. The reason is that humans are going to be able to express much more empathy and to really identify with the customer much better than any kind of automated system would. An automated system that tries to fake empathy, usually almost always fails. So, I think that they’re actually doing a good job. But it’s not specifically in the messaging context.
There are other brands that are doing things in messaging specifically. Whole Foods is a grocery store here in the United States. They recently launched a bot on Messenger. But for the most part, they’re offering it as a way of giving you recipe suggestions and things like that. So, it’s not really doing the customer service thing. It’s more about content and more about inspiration and helping you think about what you might want to plan for dinner as opposed to dealing with an order or account issue.
6. You’ve talked about the importance of monitoring those conversations, improving your own learnings, but also your AI learnings because in the end it’s all about the pleasant conversation and experience for the user. Is there maybe room for a new type of conversational designer or is the role basically wrapped up in those roles we already know?
Chris Messina: Yeah, it’s a great question and I absolutely think that there’s going to be an increased need for diversification in the field of UX and product design. There is already a number of different roles that exist that are not very commonly referred to. One is a VUI, a Voice User Interface designer. This would be someone who works on Siri or on Alexa or Cortana or any of these voice-enabled assistants.The cadence and the character and the things they say and how they say it and the way they fail and fail gracefully are all part of, you know, the Voice UI designer’s process and role.
There is also what is called a CUI, which is a Conversation User Interface designer, and that’s obviously going to be much more in the role of thinking about the way that people talk to things over chat, over text. And so that role does exist, but I think it’s very early and very emergent. I read a few stories about companies hiring English majors, or people who do writing for a living, to help them build these characters. You can imagine all of the skills that Pixar looks for in their storytellers and in their voice actors. That whole mix of storytelling comes into play in this world.
These could be novelists… just people who are used to thinking about and caring about their character and about the way that people express themselves, and they are absolutely essential to this, because you don’t really want to deal with a messenger bot that sounds so boring or so perfunctory or doesn’t vary their utterances or the way that they speak. You want a quality of interaction that feels good and is appropriate but yet isn’t too superfluous.
Poncho (the weather cat), is a really interesting example where they’ve really tried to create this engaging cat-like character that tells you the weather every day. It’s received some negative feedback, especially at the launch of Messenger, where it was one of the featured bots. But it raises this question: how do you create an engaging consistent persona in your brand through these channels as opposed to just leaving it up to the customer service rep? There have been times when I’ve talked to Virgin America’s support staff actually over Twitter direct messages, and they actually personified the brand in a pretty interesting way. So they’re actually building the brand experience on this channel as opposed to just being a very straightforward outsourced kind of support agent.
7. So, what type of personality would you give a chatbot and do you think your personal bot, MessinaBot, resembles your own personality?
Chris Messina: Well, yes. I mean, your strategy here needs to be considered in the context of your entire content marketing strategy and your branding strategy, right? It shouldn’t just be like “Oh, what does our chatbot do?”, and it’s different from everything else. It’s like “How does this fit in and what are our customers’ expectations and what’s the brand that we’re trying to convey?”
In the case of my bot, it does exhibit a lot of my personality, although I didn’t write all of the content (spoiler alert!). So, it’s a little bit more playful and a little bit more curious than I might be, but I think it’s an interesting blend of my personality mixed with a little bit more character actually. That said, the thing that is good about it is that it actually brings together a lot of the content that I publish across the web or in different contexts. So, it brings together my media pieces. It brings together comments that I leave on Product Hunt, because I spend a lot of time on Product Hunt. It brings together my Foursquare tips and it also brings together my cocktail recipes that I publish to an app called BarNotes. So, it brings together all of the content that does represent me in a way that re-contextualizes it all through me. So, even if the voice of the bot itself when you’re interacting with it is a little bit more playful than I might be in my text messaging, the content does represent me and I think that’s an interesting thing to think about and consider.
8. What type of specific transactions customers would like or maybe dislike doing with a bot? You already brought the Uber example, which is a perfect one.
Chris Messina: Yeah, I would add a couple of things that I think are interesting to look forward to. One is: how bots and automated systems can act in group environments, so: shared environments. Right now, as far as I know, you can’t add a bot to a group on Messenger. You can, however, add bots to groups on Telegram and, of course, on Slack as well. So, the ways in which bots can listen to conversations and then add useful or contextual information or be used as tools in the context of this channel is also very interesting. Right?
Let’s say you’re in Slack and you got a bunch of coworkers and you’re all on a channel and it’s time to go for lunch. Well, this is always a pain in the ass, right? So, if you’re trying to order in the US you might use an app like GrubHub or UberEats to order lunch for the group. Previously, what you used to do is have everyone download and install the app and look at the menus and then of course bring the order together and do all that. It’s just a nightmare. Whereas in the context of group conversation, you now have a shared canvas to do that collaboration work and you have bots that can interject themselves in that flow in a way that even the least technically sophisticated people can interact. The bot can basically go to each individual person and say, “Okay, here are your options. What would you like?”, and then the person can respond and then the bot tallies all that information and then reports back to the group. So, it’s incredibly valuable and incredibly useful and there’s going to be more and more of those types of group activities. Essentially multiplayer modes that previously were very, very hard to do in native apps.
The other case that I think is very interesting is going to be payments. Most of these chat platforms have some kind of payment system. Kik is interesting, because it has a virtual payment system. It’s training teenagers essentially to use Bitcoin. It’s worth thinking about how users are being slowly trained to become familiar with different behaviors. Once you’ve transacted business in the context of Messenger and paid for things or services or upgrades (like in-app purchases), that unlocks all sorts of potential because the baseline behavior has been set. You can imagine having a conversation with, let’s say Virgin, Virgin America or KLM, and in the context of the conversation, you can actually upgrade your seat just by talking to an automated system that you’re engaging with over messenger. No need to download the app anymore. You don’t need their website, which of course you have to learn how to navigate. You just feel like “Oh, I’ve got a long flight coming up, I’d like to upgrade my seat or change my seat or whatever” and the bot can do that and if there’s any fees attached to that, of course you could pay right in that channel.
9. Building bots is such a great way for people to start coding or designing products themselves. I believe there’s a huge potential here to bring people to these new technologies, people who would never have developed a native application or a website and who suddenly show interest in these kinds of things. What’s your take on this?
Chris Messina: You know, I got my start in coding HTML in 1994 but in some ways, like learning the order of these systems, learning how to take code and then output something to the screen and then make something interactive is this incredible ah-ha moment where you suddenly have the super power you never had before. I think the reason why it’s valuable to remember where a lot of this technology came from, is that it came from expert systems. It came from systems that used to have these enormous training manuals, that you´d have to learn, that were extremely esoteric, because of the nature of the hardware and the expense and all that stuff. As the cost of all these things have come down precipitously, the ability for any person to be able to contribute to the overall kind of web or publishing ecosystem that’s out there has gone up hugely and I think the interest and hunger to be able to build and do interesting things in this medium has also gone up as well.
So when it comes to building bots, I think that they are incredibly accessible and you can get up and running very, very quickly. And that, of course, many, many people will try to build some sort of bot. Now, building a good bot is actually very, very challenging and difficult, but there were a lot of really, really crappy websites back in the 90s’, right? So, even though right now, you may think that bots are fairly rudimentary and like toys, these platforms have only been open for months, not years. There’s not enough time for people to really understand how to build these things well and how to create more tooling and infrastructure.
That said, the level of sophistication of some of the tools that are out there is astronomical, if you think about TensorFlow, computer vision or other types of projects that Microsoft or Google have released in the last year. So, what’s nice is that there’s this curve of complexity that allows building bots to be quite accessible. The use of those bots is also very accessible, so that I can build something and then share it with my friends and they’ll be able to experience and encounter it, which of course creates more incentive and more motivation for me to become more sophisticated. And as I grow in my abilities, there are more and more complex tools that I can add into the mix.
So, that’s incredibly important to understand. And especially to draw a parallel: if you think about the early days of the iPhone — before there was the App Store, you had to build web apps, and those were incredibly complex and difficult to build. Eventually and over time, Apple has made it much easier to build apps for iOS with Swift and Swift Playgrounds. So, they’ve lowered the barrier of entry for building an app, but I think the conversational software is actually even easier to build with much faster pay off. So that’s really promising and I think that’s why we’ll see this proliferation of apps happen this year and next year.
10. Cool. What a great way to end this interview! We are looking forward to what we are going to see on these new platforms. Do you feel like we should touch anything else? Or anything you think I should have asked you?
Chris Messina: I guess the one thing I would add, is to really pay attention to the number of platforms that have been released this year alone. There’s strong evidence that there is a land grab underway, and it’s unclear who exactly the winners are going to be. Watching that space is going to be extremely interesting, but taking a long view is equally important, right? There will be a bunch of people who see this as an opportunity to trick the system and ruin the experience for the rest of us. However, brands and companies and builders who think long-term are aware that messaging is here to stay. The ease of use and the accessibility of computing devices will open the door for the other 4 billion people who are yet to come online, and who will be probably be using messaging when they do. That’s the thing to think about. That’s the opportunity.
Free eBook Download!
It’s not about “What can I get done in 2017 to spam my users and get a little bit of uplift?” So, I just want to keep reiterating that, because I think there’s an opportunity to really ruin this for us, if we don’t think about that and hold to that long-term perspective on how we can really do things. Think about that. If my story started in 1968, that’s almost 50 years ago. It’s taken nearly 50 years to get to where we are. Google is almost 20 years old. So, if you think about what’s going to happen in the next 5, 10, 20 years, there will be another Google, there’ll be another Facebook, there’ll be another Snapchat. And so if someone aspires to build that, I definitely think that taking that long view is the way to do it.
About Chris Messina
Chris Messina is a friend to startups, inventor of the hashtag (original proposal), former Googler, and proud participant in the open source/open web communities. He co-founded the BarCamp and coworking communities, and on top of all that, he is a designer, thinker, writer, Twitterer, and product hunter.
Want to learn more about the future of chatbots and how they are changing online interaction for both individuals and businesses? Read the interviews on our blog or download our eBook.
Barbara Ondrisek, Like a Hipster: “When You Stop Developing Software, It Instantly Dies”
Thomas Schranz, Blossom: “Chatbots Transform the Customer Support Experience in Every Industry”
Matt Schlicht, Chatbots Magazine: “Bots Represent the Apps of the Messaging Platform World”
Bernhard Hauser, oratio: “Customer Service Teams Can Coexist with Chatbots”