Bot Discovery & Virality Lessons From Kik CEO Ted Livingston

We recently caught up with Ted Livingston, CEO of Kik, at the Talkabot Conference in Austin, TX and got his key lessons for bot discovery and virality.

Kik started working on bots two years before they suddenly became cool in 2016. The platform is the first to pioneer many social bot discovery and virality features for bot developers that don’t exist on any other messaging app. We were keen on getting Ted’s insights into what’s worked on Kik’s platform that can be applied to bots everywhere.

Be sure to read the interview to the end, as Ted covers fascinating topics such as:

• His top three tips for bot makers to be successful
• How bots go viral on Kik and how you can do the same
• Kik’s platform-specific discovery features like invites, mentions, and concierge bots
• Whether artificial intelligence is necessary for bots
• Which big tech company will win the bot wars

TOPBOTS:

We work with many brands and businesses building bots and they all share two common issues:

1) Onboarding. People don’t know what bots are and they get confused.

2) Discovery. Everyone complains that bot discovery is a challenge. What do you think has been most useful for solving these problems, especially bot discovery?

Ted Livingston:

We want to create bot discovery on Kik that’s sustainable over time. One day there will be millions of bots in our bot shop. Just like mobile today, you won’t be able to get discovered. Nobody will get your bot.

TOPBOTS:

Paid acquisition is the only way to acquire users on mobile right now.

Ted Livingston:

Paid acquisition will become the only way and that favors the big guys. The biggest advantage of bots is there is no friction for trying them. You don’t have to download an app, you don’t have to create a new account, you don’t have to learn a new interface.

The question we asked ourselves as: how can we create a platform to best leverage that for bots to go viral on their own merit? You might get your first couple users through the bot shop or paid acquisition. But from there, if you actually build something good, you can go viral through Kik no matter who you are.

That’s been our perspective on bot discovery since the beginning. Two things we do that I really like are mentions and invites. Mentions is the ability to bring a bot into a conversation, a chat or a group chat. Bots can also use mentions to bring each other into conversations.

TOPBOTS:

Like how the Sephora bot calls in the BeautyTube bot to show editorial makeup tutorials?

Ted Livingston:

Yes. Fashion and beauty is a killer category for bots, whether you are online getting fashion and beauty tips or offline when you’re in an actual store.

The way mentions makes bots go viral is like this: you start with one user. This user calls your bot up in a group chat and now you have 10 users. Those 10 users call up your bot in different group chats and suddenly you have 100 users and then 1000 users.

The second social bot discovery feature is invites. Invites allow bots to go viral through Kik the same way that Facebook went viral through email in the early days. When Facebook launched, someone would tag you in a photo and you’d get an email notification. You’d say to yourself “What’s this Facebook thing? What’s this photo?” and go sign up to see the photo. Then you’d tag a bunch of friends not yet on Facebook and they’d repeat the same experience.

Invites is the exact same feature but for bots in chat.

TOPBOTS:

Can you give me a discrete example of this?

Ted Livingston:

Say you’re playing a game like Zombie Survival which asks you which people you want to bring onto your team. You pick 10 people you’re chatting with who have not heard of the bot, but these invites give the bot permission to reach out and tell them you want them on your Zombie Survival team and ask “Do you want to play?”

TOPBOTS:

That makes tons of sense.

Ted Livingston:

Invites can go to people you want to play games with, people you want fashion advice from. One of the unique things with Kik as a platform is that Kik is the only messenger where you get complete control of your identity. Your identity is not tied to phone number, a social profile, or your real life story. Your identity is just your Kik user name.

So, we can be much more aggressive in having bots spread through the network because they only access your Kik username. We don’t give bots your phone number or access to your social profile. The key part of Kik is you can connect with new people whether they are real humans or bots.

TOPBOTS:

So Kik’s anonymity leads to better distribution and bot discovery? Because you’re not worried about privacy violations?

Ted Livingston:

A big reason teens use Kik is they are spending more and more time on the online communities. 2/3 of teens have made friends online. 1/3 of teens who meet outside of school exclusively hang out online.

So when teens meet people online, they want a messenger that gives them complete control of their identity and that’s Kik. A new bot is no different than a new person. Like somebody reaching out and saying “Hey, I know you from such and such a place, let’s be friends.”

In a recent example, we had 2 people at Kik who built a bot called Roll in their spare time. they launched on Kik and they didn’t tell me.

TOPBOTS:

Time spent well at work!

Ted Livingston:

Well, it wasn’t even at work. Roll was built outside of work. And the bot went from 0 to half a million users in a few weeks.

The reason why Roll is interesting is that the bot got no special treatment, was built in somebody’s spare time, and all they did for bot discovery was put Roll in the bot shop.

TOPBOTS:

How long did Roll take to get from 0 to half a million?

Ted Livingston:

I don’t know the exact time but a matter of weeks. 95% of their chatters were organic, meaning they didn’t come from the bot shop.

TOPBOTS:

Wow. What exactly does Roll do?

Ted Livingston:

So they built a bot that helps you decide things in group situations.

TOPBOTS:

Oh that’s fantastic. Teenagers need that.

Ted Livingston:

Roll just spread like 95% organic and only 5% from the bot shop.

Ted Livingston:

People are still experimenting with how bots should work, how they should design the experience, how they should grow. I think the big distraction right now is people really focus on natural language processing. They emphasize that you will chat with a bot.

TOPBOTS:

I noticed you are very good at not doing that.

Ted Livingston:

We tried that a year and a half ago and people go “I don’t know who this person is, I don’t know what I should say to them”. They see that blank cursor and literally they swear at the bot. Then they never chat with it again.

That’s why we built suggested responses. People are trying to catch up to that now, having a very guided experience but layering in rich elements like full web games, interactive experiences.

TOPBOTS:

Right. The mini app concept.

Ted Livingston:

Exactly. Exactly.

TOPBOTS:

I tried really hard to chat with some of your bots just to see the reactions. You actually make it hard to write a real message. You block the user input which I think is quite clever. Clearly you’ve been working on it longer than everyone else.

Ted Livingston:

Yes. Clearly we have been doing this for a while. Step one is helping everybody and working through challenges together of “How do you build a great bot?”.

Once we figure that out, the second challenge will be “How do you grow a great bot?”. That’s where we work on bot plaforms and know why we created certain features and how they should work. We can get great examples like Roll where we can say “Yup, it works. You can get half a million users in a matter of weeks.” That’s part of why I’m here, to work with the ecosystem and help people with bot discovery.

TOPBOTS:

Let’s talk about this whole bot-to-bot discovery, bots making friends, bots calling in “concierge bots” like Sephora does with BeautyTube. I know that this is relatively new. What have you seen about that specific feature of “deep-linking” between bots driving growth or discovery?

Ted Livingston:

We built the feature in because we thought it would be very powerful over time. One of the key lessons about how you build a great bot is it does one thing really well. That’s what we tell the developers.

The question that comes up is “What if it needs to do this or what if it needs to do that?”. What used to be really simple becomes big and complex.

That’s when we said “What if we had a bunch of bots that are all individually really good at one thing. Then when they needed to, they can collaborate to get things done.” This is similar to how the web started. You have a website for this and website for that, but they can call each other’s APIs to get stuff done for the user.

If it works in web in the browser, how can we make it work in chat? So that’s what we built into the platform, but it works similarly to the invites feature I mentioned before. We are still early in seeing developers take advantage of that.

TOPBOTS:

I see. What are some things you want bot developers to try? What do you think would actually help with bot discovery and growth using this bot-to-bot feature?

Ted Livingston:

So I think the bot-to-bot feature is still a little bit on the horizon. You have to build something great first, you have to grow it, and then you can start thinking about how your bot fits into a bigger picture with other bots.

Right now my 3 top pieces of advice for bot developers are:

1) Make sure your bot is really simple and does one thing really well

2) Think about how it will grow. How can you get a user to say “Yeah, I want to do something with these five friends” which then, boom, gives the bot permission to reach out.

3) Always make sure there are suggested responses on the screen. If you leave the user on that blinking cursor, they are lost.

TOPBOTS:

Do you have any numbers around that? What’s the frequency with which bots on Kik use suggested responses? How do they outperform bots that don’t?

Ted Livingston:

Basically 100% of the bots in the bots shop have suggested responses.

The first version of our bot platform 2 years ago was based on a key word engine. So as a brand you set up your bot by mapping user keywords to specific bot responses.

TOPBOTS:

So like what Assist does. Enumerate menu options and ask the user to type in a specific number for a specific response. Feels like Microsoft DOS.

Ted Livingston:

Yeah. We did a bunch of user tests and users would just stare at the screen. To get any reaction, bots had to be very explicit like that. To do this, say one. To do that, say two.

That’s when we decided the bot will always tell you what it could do and then give you the menu. Why don’t we just have the menu be the keyboard itself?

TOPBOTS:

What I liked about Kik’s suggested responses is that you are not able to easily type. On Facebook, you can get quick replies but you can still type and that causes errors.

Ted Livingston:

We tried that as well but developers say “I don’t want to support this case because the users just get lost.” Every developer would react to unknown user input with “I don’t understand what you’re saying” and just repeat the question. Why are we setting up the both user and developer for failure?

TOPBOTS:

One of the best ways to eliminate error is just to not allow that path to occur. Instead of a user tapping a button and you showing an error message, you simply disable the button so no action is possible.

Ted Livingston:

This is one of the big key themes. A lot of people are looking at chatbots and saying “Oh, it’s all about chat, it’s all the natural language processing. It must be about actually talking,” when really it’s not that at all. Bots are very simply just a better way to deliver a software experience. Bots are just an app disguised as a chat.

TOPBOTS:

Lightweight “almost” apps.

Ted Livingston:

Yes, but over time they will get richer and richer and richer. Right now, you can have full web come up so you can embed a full game in your chat. You can be talking with the bot and it tells you “Ok, it’s time, we are going on a raid to this place,” you click start and, boom, you are in a full game as if you were in a native app.

TOPBOTS:

Facebook just announced a bunch of templates, like the list template. What are you thinking about from the UI template side of things that you think would help bot discovery and growth?

Ted Livingston:

So I can come back to that. We talk to developers and they are saying they see 10–100x more chatters on Kik then they see on Facebook. Yet Facebook messenger has a billion monthly active users. We are nowhere close to that. We are fraction of that. And yet 10–100x more chatters on Kik is I think due to the bot discovery options.

Part of that is the demographic, teenagers have more time. But part of that I think is the way identity works. On Kik, you are used to reaching out and chatting with new people. It’s a core reason to use Kik. Whereas on Facebook you only use it to talk to old college friends or distant family now and then. Everybody else you text.

TOPBOTS:

Interesting platform positioning.

Obviously the big question is who is going to win the bot wars, right? Now everybody is in. Literally everybody. There is not a single tech titan not in the bot wars. Because you have fewer users, people may perceive Kik as being at a disadvantage. What’s your strategy knowing Facebook’s strategy and their limitations? Knowing Google’s strategy and their limitations?

How do you position Kik?

Ted Livingston:

Everybody is the bot wars and at the same time nobody is in the bot wars. I think actually the only “tech titan”, to use your words, that’s in the bot race is Facebook. Amazon, Google, Microsoft — they’ve put a ton of investment in terms of money and engineers into natural language processing and AI. They are looking at bots as a way to deliver that technology to consumers.

But what we are seeing is consumer don’t want that. The key thing you need to be in this race is you need a popular messenger. Neither Google nor Microsoft nor Amazon has a popular messenger.

TOPBOTS:

Right. They are more invested on architecting like the infrastructure for conversational AI.

Ted Livingston:

They are hoping that AI becomes the key component to winning this big race. If it is, the number of people that have the resources and the engineers to build that technology is just them. So, of course you are telling yourself the story that “AI is gonna be the key thing that lets us win the race. We are one of a handful companies that has the resources to do AI properly. Therefore we will be one of a handful companies that will win this very important race.” Meanwhile, Facebook is sort of like quietly snickering in the background, thinking “These guys are totally missing it.”

Chat is going to be the next big thing. It’s the once a decade platform. We’ve already seen that in China. Everybody thinks the thing that you need to compete in that race is amazing AI technology, but in reality you just need a popular messenger and guess what? The only popular messengers for US consumers are Kik, Facebook Messenger, WhatsApp. Facebook Messenger and Whatsapp are much bigger then Kik. So let’s just let them keep going.

TOPBOTS:

Let’s go back to the idea of designing a bot that’s actually gonna become popular. Now obviously there are certain things that are more viral than others. A video of someone BASE jumping off of a mountain in a wingsuit is going to be more popular then a video of you making oatmeal.

If you are advising a bot maker that is trying to get big on Kik, what would be your advice in thinking about how to take advantage of these mentions and these invites? For utilitarian bots, there’s a natural limit to how viral you can be.

Ted Livingston:

My advice would be to find a way to require your users to pull in their friends. What you said about going viral is true. Some things will go more viral then others but the big advantage of bots is that the barrier and friction for things to go viral is an order of magnitude lower than any platform before.

So, for example, something could go viral if you say “Hey, if you want to do XYZ with our bot, you need to invite 5 friends.” Boom, that bot gets to reach out to those 5 people, then they repeat. It could be that simple. So if you can demonstrate just a bit of value, I am convinced that pretty much any bot can find a way, at some level, to be viral.

TOPBOTS:

Even as an arbitrary invite gate?

Ted Livingston:

Now the best bots will find a way to do it in a way that is compelling and natural. But if you just start with that: what feature could we gate with an invitation to your friends? Then, how can we make that invite process simple, intuitive, and natural for the user.

TOPBOTS:

If you have a premium feature, users can unlock it by either paying or inviting their friends. That’s what Dropbox did. Pay to get more space or invite friends to get more space.

Ted Livingston:

Exactly.

WANT MORE FROM TED?

Watch Ted’s Talkabot fireside chat where he goes into greater depth about AI for bots, what we can learn from WeChat, the future of the bot economy, and what bot makers should do to maximize success.