How bots, messengers and apps will define 2019

On the heels of our 2018 retrospective, we decided to take a look at what product themes and technology are likely to define the year ahead.

I’m joined once again by Paul Adams (VP of Product) and Emmet Connolly (Director of Product Design) for a conversation that covers everything from bots and automation to the state of messaging.

Short on time? Here are five quick takeaways:

  1. Bots are finally coming of age. As the world’s population grows – and that population migrates online – bots are increasingly being used to address the challenge of helping businesses scale. But instead of focusing exclusively on algorithms, the next frontier is making those bots accessible and user-friendly.
  2. Chatbots shouldn’t just ape web forms. Instead, they should be deployed when there are more complicated if-this-then-that scenarios that would be difficult for users to guide themselves through.
  3. One of the big success stories for Intercom in 2018 was our App Store. Looking at the year ahead, we expect that apps will no longer be a destination in and of themselves; rather, they’ll exist anywhere and everywhere to add native value for the user.
  4. Sometimes the biggest gains come from revisiting and improving old products, rather than simply introducing new ones. That’s why we’re redesigning our onboarding process from the ground up, with the aim of personalizing the experience for every single Intercom customer.
  5. To get inspired for the year ahead, we’re turning to books. Emmet’s reading The Design of Childhood by Alexandra Lange, which looks at how we create and recreate spaces for our kids. Paul’s engrossed in Principles by Ray Dalio, which follows an investment banker’s successes, failures and future redemption guided by a set of ideals.

If you enjoy our conversation, check out more episodes of our podcast. You can subscribe on iTunes, stream on Spotify or grab the RSS feed in your player of choice. What follows is a lightly edited transcript of the episode.

Des: We’re going to talk entirely about 2019, and this will be mostly looking forward. But before that, we’ll just briefly look back to October of 2018, when we had this event called The Next Chapter in San Francisco. At that event, Eoghan McCabe, our CEO, unveiled the next chapter of Intercom. Loosely speaking, it was all about our Messenger, all about bots and automation, and all about apps on the Intercom platform – and how we really see that as the next area that we’re going to lean into.

Paul, you’ve always hated bots. What changed that?

Paul: I fell in love with a bot. So the landscape’s changed, for sure. Bots are happening, and bots are coming. I’m convinced of it. I think bots are going to be a massive success as a technology.

Des: Just for the sake of our listeners, and maybe assuming you’re on the same page, what do we mean by bot here?

Paul: You can probably break it down into multiple ways to describe them. For me, a bot is simply a piece of simple computer program that runs some logic that can do things for you: “if this, then that.” Then you’ve got chatbots.

Des: “If this, say that,” basically.

Paul: Yeah, exactly. Or if this happens, the bot pops in and does something else. And then you can get smarter with machine learning and stuff.

Bots are coming of age

Custom Bots – Hype Cycle for Custom Bots

Des: What has changed in your perception or your belief that they’re going to take over and rule the world or be a significant part of our future?

Paul: I’ll try and keep this short, because I could be here for an hour talking about this. The basic gist of it is that I’ve always believed that bots were useful. Humans are great at some things and bots are great at other things. Bots are great at things that are suitable for computer calculations, like when your next bill is due. Or showing you your bank balance. A bot can answer immediately, whereas a human has to go look it up. So a bot’s just better for that type of thing. Also, if a person is sitting there answering people’s calls about their bank balance eight hours a day, they would go out of their mind with boredom. There are just times when bots are great. And there are times when humans are great, like when someone’s angry or sad or needs a hug or some kind of relief.

Des: Or has a really complicated inquiry.

“There are all these prospective customers, but businesses can’t scale at the same rate, so they need automation”

Paul: Right. There are really fuzzy end points to it. I think what happened was – with bots specifically, through 2016 and 2017 – a lot of it was on Facebook Messenger. People started to think that bots were the answer to the internet scale challenge they have. And the internet scale challenge is very simple, which is, the human population is growing at a crazy rate. More and more of those people are going online, therefore the internet population is growing at a crazy rate. There are all these prospective customers, but businesses can’t scale at the same rate, so they need automation. Bots were heralded as the answer, the savior. But they were executed in a pretty naïve way and just used for everything. All these things that humans should have been doing, bots were doing. They got a bad rep, but they’ve come through the other end, I think.

Des: Do you think we’ve scoped down the problem space to the point that now there’s a much neater fit between things we try to apply a bot to and the suitability of a bot for a problem?

Paul: Yeah, for sure. But then the second thing is that people have realized what bots are and are not good for. I remember two years ago, it was all about AI bots. And you’re like, “What does that even mean?” And now, if someone says they have a new AI bot, you think that sounds like a bunch of crap. Whereas if I tell you there’s a bot that can book health insurance, that sounds like it could be useful.

Des: Emmet, I’ll come to you on this question, but I have to offer one of my own thoughts, which is something that has frustrated me for a while. It’s the idea of taking a web form (which includes labels and text inputs) and let’s say a “Contact us” form, and simply putting speech bubbles around both sides of that. Instead of saying “name” with an input field, the bot would ask, “What’s your name?” And you would say Barry. And it would say, “Hi Barry, what’s your address?” It was never clear to me what the hell the value prop there was, right? We already had web forms, but now we have them with a blue gradient around them. Fantastic.

I often thought that you end up in this world where bots are competing with UI. To take Paul’s example, a bot that tells you your bank balance. The thing that’s even better than a bot telling your bank balance is just a screenshot of your bank balance, you know? How do you think about when we should use UI versus when a bot has something net new to add?

Emmet: I don’t think there’s some magical thing about bots as distinct from most other forms of automation. In some sense, as Paul was saying, a bot is essentially a simple computer program. And there are a lot of cases historically where automation comes along and takes over a job that humans would have done. There are new patterns and new framings – like the chatbot framing – that come along and nibble away or eat into some new use case. But fundamentally, you have to think about what you’re trying to achieve and what’s a good mode in which to achieve it.

Take your bank balance example, there. Banks used to have bank tellers, right? You’d go to a bank if you wanted to get a bit of money out, and you’d queue up and wait 15 minutes to get to a person who would hand you your money. Then ATMs came along, and the people who want to just take out some cash can go to the ATM and do that very low level transactional thing because it’s better than waiting a long time. But if your account is in overdraft, or you want to apply for a mortgage, you still need to go and talk to a human. So I think even within one domain, there are different tasks that are suitable to be automated versus those that are not suitable to be automated.

“You’re really looking for a fit between what you’re trying to achieve and the technology that’s available”

Another interesting side note, by the way, is that there’s a lot of concern about the societal impact of automation and people losing their jobs. When ATMs were introduced, which I think was back in the early ’70s, it didn’t put bank tellers out of work. More bank tellers – across America at least, which is where I read about this – were employed because it meant you could have more banks available to more people in smaller towns, smaller branches, right? Often I think about this stuff in terms of automation, which is to say there’s this menial low-level task I want to replace – but a lot of this stuff is also around augmentation. There’s still a human in the mix, but how can we help them be a bit better at their job? Maybe by suggesting what they should reply with and things like that.

There’s a lot of nuance, and you have to think about what you’re really trying to achieve and whether this new tool that’s in the toolbox that’s available to everyone fits. For us, it fits. I don’t actually think we even used to be down on bots, but now we think bots are amazing. Bots happens to be very squarely aligned with what we’re doing. We have a messenger, and people often contact businesses using Intercom with a very simple question. You’re really looking for a fit between what you’re trying to achieve and the technology that’s available.

Des: It makes sense. The areas where I think chatbots obviously have a clear value are where the user has no intention of ever learning your interface whatsoever, but they do know how instant messaging works. That’s a very clear sort of value prop.

I still see bots in so many cases asking, “What’s your company size, and what’s your budget?” And it’s basically just a web form masquerading as an AI-driven chatbot. But where they can actually beat a traditional web form is when you’re taking somebody down various different paths based on their answers. The form version of that is: “If no, skip to question 7A. If yes, skip to question 8B.” I think bots can do that piece directionally better than forms. But then we still have that debate about interface and whether an adaptive UI might beat it yet again. It will be interesting – if people ever test it – to see what will come out on top.

Paul: I think one of the advantages of bots is that they’re easier and faster to build. You know, with a good bot builder you can build different flows and endpoints, whereas trying to build a modular UI is way harder.

Des: That’s true, and they’re probably more accessible to a lot more people as well, with the right bot-building tool like Intercom’s, for example. In practice, any marketer who wants to put a bot live on any page on their site can just drag and drop some stuff together and basically ask the questions they want, and then trigger the thing – whether it’s to register for a webinar or to start a video call or whatever makes sense. So I do agree, there.

Intercom Custom Bots

Intercom’s custom bot builder

Des: You hinted at automation versus augmentation. In general, we’re seeing automation appear everywhere, and all the menial tasks of working online, behaving online and acting online are starting to fall away. When you think of this beyond chatbots, where do you see this happening? I’d love to hear some more about this sort of difference between automation and augmentation.

Paul: It’s a moving boundary, right? Back 40 years ago, the frontier of this was bank tellers doling out cash. It seems like bus drivers and car drivers are the current frontier, and I’m sure that will be different again in several years time. Last year, as we dove into this more, I thought a lot about the opportunities for democratizing access to actually build these things. Up until quite recently, if you wanted to build an ML system, you needed to have a hardcore engineer who understood ML and used TensorFlow or one of these products that were very inaccessible to most product people.

One of the things we built last year was our Resolution Bot, which answers simple questions if it feels like it knows the answer and is based on machine learning. But we didn’t want the people using Intercom to have to train the system in a very technical way, so we built this UI that allows you to go in and train it on what your questions and answers are.

We looked around and didn’t really see anything, apart from a lot of prior arcs and existing systems that were what we really came to think of as content management systems for ML or UIs for training an ML. So I’m really excited about that idea and about those things becoming a little less opaque to even very technical people like us, who may not be active computer scientists.

Another one of my favorite tools from last year is called by Mike Matas, who was an early Apple employee. It’s a drag-and-drop system where you can build your own AI systems. To me, that definitely feels like a frontier, because a lot of this stuff is just still really hard to build manually, you know?

Apps everywhere

Intercom’s App Store

Des: We said earlier we’re going to talk about messengers, bots and apps. One of the big success stories for Intercom last year was our App Store. There are 100-odd apps on it. You can now pay for something through the Stripe app, or you can order a product through the Shopify app, or any of those types of things.

Paul, I feel like it was like four years ago that you talked about The End of Apps As We Know Them. It was the idea that apps would no longer really be a destination, but they’d really be able to exist anywhere. That certainly is what’s played out for us and in Intercom, but obviously we’ve seen it play out elsewhere: inside Facebook, inside Facebook Messenger, how Alexa can talk to apps and all that sort of stuff. The idea of apps being singular products has faded away a bit. Do you think that will keep going in 2019, or have we seen as much as you expect here?

Paul: Yes, and it’s interesting to watch it play out. I remember when I wrote that blog post in 2014, Google Now had just been announced. And I saw Google Now as a pretty amazing thing because it was like a feed of apps, basically. I had just come from Facebook, which is also this modular ecosystem where apps are a core part of the experience.

“I thought at the time that apps would no longer be destinations that you would drive traffic to”

I thought at the time that apps would no longer be destinations that you would drive traffic to. And instead, they would start to appear all over the place. “The Internet of Things” was also this quite cool thing at the time. I guess the surprise for me potentially was it just took so long to play out. I don’t really know why that was; maybe a lot of the technology had to mature, and the distribution networks felt the apps had to mature.

Des: Probably the product mindset of everyone who’s working on these things, in a sense. There are all sorts of APIs and shit that need to get built, but you need to prioritize that work.

Paul: That’s very true – even apps like Stripe. I don’t know when Stripe was founded, but in 2014 they weren’t very mature. So over that time, the whole landscape has matured, as you said. I do think it will continue. We’re going big on this idea at Intercom. We have an App Store, and it’s still early days, but it’s doing really well. We’re going to invest and invest and invest in the developer side of it and help external developers get value from building an app in the Intercom ecosystem.

I still wonder about other places, like media. Google Now looks a lot like it did in 2014. It still looks like a lot of news websites and cards, so I wonder where all the apps are. Gmail announced this app framework last year. Again, that’s something I would have thought in 2014 was happening imminently. But it exists now. If that app ecosystem takes off, I think it’s just good for everyone.

Des: They have, what, 800 million users? You have to believe them.

Paul: It speaks to the difficulty in designing a platform, though. There are just so many dynamics at play. You’ve a multi-sided marketplace of users and builders and you need to get all the incentives right for everybody for it to work. So it’s an extremely tricky community management problem, as well as just building the right product.

Des: I actually worried about that a bit, because when we were building our OS, I was always thinking, “I can see why we want this.” For us, it was about extending the capabilities of our Messenger so that we didn’t have to build literally everything everyone would ever want, because obviously we had 30,000-odd customers at the time. But it was still a hard-fought battle to get those products out there, right?

Paul: It might be interesting to distinguish between the consumer side and the enterprise side. With consumer software, you tend to get a very small number of big, big winners – like on the Facebook ecosystem, you know, there are very small numbers of big winners.

Des: Like, have you been bought out, right?

Paul: Spotify grew substantially via Facebook. I think one of the causes of Pinterest’s growth was Facebook. So you’ve got these big winners, and then just hundreds and thousands of losers of it, where it didn’t work out. In B2B software, where businesses are selling through to businesses like Intercom does, you might see different dynamics because every business is unique.

“We tend to think of these apps as the last mile, the last piece of the puzzle”

Like you said, we can’t build everything people want, so we tend to think of these apps as the last mile, the last piece of the puzzle. We can get you most of the way there, and then these apps you can use in our ecosystem mean you can get the full job done. And because there’s a wider pool of types of usage, people will tend to use a lot more of them.

Des: To make this tangible, Emmet, what are your favorite uses for our Messenger platform?

Emmet: When I think about when we launched the App Store and truly opened up the platform, one of the most exciting things for me was to see lots of things we had debated building but either decided not to or just hadn’t managed to prioritize that stage. Some of it was big stuff; for example, Aircall built an app that allows you do video calls inside our Messenger, which was a thing we had talked about doing for a long, long time. But there was even narrower stuff like making a little app to do NPS, which again was something we might have done. But especially if you’re more of a long-tail customer, there are really specific things Intercom never could have justified building.

Des: What’s that screen sharing one, Loom? Where a customer wants to share screens with customer support so they can de-bug their thing. And of course, sure the customer wants that, but that’s not a common request, and now we can actually point you to a perfect working solution through Loom.

Emmet: That’s probably a killer use case, and a great use of Intercom for a bunch of people out there.

Paul: And quite honestly, Loom will execute it better than we will.

Des: The other one I really liked was the translation one, because again, that was something that we’ve been asked about forever – to help two people who don’t share any language in common speak a common language, simply by translating on both sides.

Again, that’s something we always knew was going to be possible when you hook up Google Translate’s API. But it never got to the top of our prioritization stack. What’s really fascinating for me is that when you actually create these platforms well and make them enticing to build for, prioritization takes care of itself. While it might not be our priority, if you’re Babel, it’s probably your priority to be on one of the most popular messengers for businesses out there, you know?

Emmet: Back to what Paul was saying about the winner-takes-all type dynamic in consumer products (and then B2B being a bit more about distributors), it feels to me like there are a lot more opportunities for a mutual victory to happen. You know it’s good for Intercom because it strengthens the platform, and it’s good for, let’s say Aircall, because they get greater reach, and it’s obviously good for the customers of both of those products, as well.

So I think that kind of dynamic emerges more easily in the B2B space where people are paying for the products, and therefore both those businesses can win by providing a better service to the same customer. Maybe there’s more of a route to success for businesses like ours that charge money for their products.

What’s ahead in 2019

Des: We’re sitting here in January, heads down, finalizing our plans for the year. When we think of the next 12 months for product and engineering teams, what are some of the high-level themes you’re both focusing on?

Paul: Yeah, it’s actually fascinating for me. I’m smiling here. We’ve talked about bots and apps and the vision and all that stuff. But maybe the thing I most excited about right now, is helping our new and existing customers get more value out of Intercom, where it’s basically onboarding over a long period of time. There’s your first use, and then there’s ongoing use. Last year, we shipped more than 140 customer-facing features, changes, brand new products, the whole thing. And I guess we just didn’t invest in onboarding as much as we might have.

Des: It’s hard to keep pace, really.

“The goal is that every single Intercom customer – new customers, existing customers – will have a personalized onboarding experience”

Paul: It’s hard to keep pace, exactly. So we’re in the process of completely redesigning onboarding from the ground up. It’s this whole new modular system, where ultimately the goal is that every single Intercom customer – new customers, existing customers – will have a personalized onboarding experience where we’ll offer them the next thing to do. “Did you do A, B and C? Try D.” Or, “Hey, you’re all the way down Q, R, S, T – you might want to try something else.” I’m phenomenally excited about that. I think it’s going to have a huge impact on our business.

Des: I think this is true for every product at some point: the best, biggest value comes from revisiting old product rather than net new product, and it certainly seems to be the case here.

Paul: I’m definitely excited about a bunch of new product that we’re building, but I would imagine our marketing team wouldn’t consider this the ideal forum to . . .

Des: You heard it here first: new product!

Paul: A big thing for us outside of design has been going global. In years past, a lot of our R&D effort has been based in Dublin. We have R&D in San Francisco and London now, so a big opportunity we’re actively working on is thinking about how to build a global team. To what degree should we be one coherent team versus the opportunity to break into loosely coupled sub-teams that are pointed in the right direction but not shackled to one another?

Des: But locally specialized if need be.

Paul: Exactly. There are a lot of exciting opportunities there for us to really scale up in a way we weren’t been able to when we were just in one location.


View this post on Instagram


On this #TBT we’re reminiscing about the opening of our London office last fall 🎉

A post shared by Intercom (@intercom) on

Des: As you both look ahead to 2019, knowing that we’re probably going to hire a lot of people in all three of our product and engineering offices, what are the organizational challenges you’re going to keep a close eye on? Is there stuff you anticipate might break or go off the rails unless you pay a lot of attention?

Paul: As we’ve grown, trying to scale an appropriate hierarchy is an ongoing challenge. I mean that in a good way. As you add more people, you need more layers. And what I mean by appropriately hierarchal is that it’s not my job or Emmet’s job or your job, Des, to be in the weeds on any of the work. We need to continue to design our process and operations so that we’re not and so that we have local leadership and empowerment within the teams, within the guidelines of our strategy. Our job increasingly is to deliver strategy alongside our other leaders in the company. Then, each layer down will take their part of it and execute it really well and make sure they’re connected to the pieces they should be connected to.

Emmet: I don’t know if there’s some specific “Oh my god, this one problem is going to screw us in the year ahead” thing in my head, but I think about this in terms of what’s often referred to as design thinking.

“Design thinking is the ability for organizations to take a more iterative approach in how they organize themselves”

Design thinking is the ability for organizations to take a more iterative approach in how they organize themselves. And given that Des and Eoghan, our co-founders, and Paul are designers, I think without explicitly calling it “design thinking,” we have started to realize that we have always taken a very iterative approach to how we run the company and how we run teams.

At this stage, I have a certain degree of blind faith that something will go wrong in the year ahead and that we will have the ability to recognize it and self-correct, because we’re constantly revisiting our process, and because we’re constantly going to the teams and asking them what’s starting to break for them. It’s hard to know what specifically will break, but like I said, I have a high degree of confidence in our ability to self-heal and work around those things fairly quickly.

Inspiration for the year ahead

Des: Last question, different topic. To give our readers some inspiration for 2019, what are you both reading?

Emmet: I am reading the internet. I do that quite often. In terms of books, I’m reading a book called The Design of Childhood by Alexandra Lange, who is The New Yorker’s architecture critic. This makes me sound super smart, though.

Image result for The Design of Childhood by Alexandra Lange

It’s about the design of childhood. It’s about the objects that occupy small children’s lives and how they’re designed and how those designs affect them. What’s really interesting about it is that she starts with building blocks and small toys that children are first exposed to, and then she looks at how houses are often “childproofed,” as it’s called, and how we redesign the house around the child. Then their street, their neighborhood and finally, their city. It’s just like a Russian nesting doll approach.

“That nested approach refers to how we think about and talk about system design”

Structurally, it’s very similar to a book that we refer to quite a bit called A Pattern Language by Christopher Alexander, which is also an architecture book. But it’s about how the design of things starts with the detail, and then you consider the context around us, like the objects in a room and so on outwards. By the way, this Design of Childhood book is interesting to me because I have kids, but also that nested approach refers to how we think about and talk about system design, which is quite an abstract thing to grok. Because architecture is a spatial thing, it’s easy to relate to, and it really helps orient your mind in that direction.

Des: We’ll be sure to link it up in the show notes. Paul, what are you reading these days?

Paul: The best book I read last year that I am still referring back to is called Principles by Ray Dalio. It’s brilliant. I highly recommend it. It’s a strange book in that the first third or so is Ray recounting his life. He’s in finance and is a very successful investment banker who started his own company, went bust and went back up again. So the first third of it is an autobiography, but then it all makes sense as he plays out the principles he derived from his life.

And I realized while I was reading this book that we’ve run Intercom on principles. We never really explicitly said it, but it came up here and there, and we started saying it a lot more. And we’ve started writing down principles. It’s been a huge success, so I’d highly recommend the book.

Des: Awesome. Anything else?

Paul: There’s another fan-favorite coming back to me, which is the Michael Porter paper called What Is Strategy?

Des: That’s an oldie.

Paul: It’s an oldie, I think from the ’70s? It basically explains what strategy is, which is always useful. It just talks about how you need to build a connected strategy. It inherently involves trade-offs: you need to decide to do one thing versus another, and it’s just a brilliant paper. And again, you introduced it to me, Des, a number of years ago, and we’ve enacted this in Intercom. We aren’t actually talking directly with the paper, but now as we’ve grown and scaled I’ve started realizing that I’m starting to send this paper to people and talk about how Intercom is playing at strategy in the Michael Porter version of the world. So that’s another goldie I think everyone should have a look at.

Des: I’m just happy that you read things I send you. Paul, Emmet, thank you very much, and we’ll look forward to chatting to our listeners again at the end of the year or maybe sooner.

Paul: Sounds great.

Emmet: Cheers, Des.