Skip to main content [aditude-amp id="stickyleaderboard" targeting='{"env":"staging","page_type":"article","post_id":2026231,"post_type":"guest","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"bots,","session":"C"}']
Guest

How this selfie-enhancing chatbot harnessed empathy to grow like crazy

Messaging apps draw more users than smartphone games, utilities, and even the most popular social networking apps.

That’s exciting news for chatbot companies that want to build something instantly relevant and useful to millions — that is, you know the user base already exists, and it’s ready and waiting. But how do you get their attention in a way that’s meaningful? There are no well-established “bot store” corollaries to the App Store — where everyone flocks to one portal using the same phone — and contemporary smartphone users are not yet well-accustomed to chatbot discovery. So what does it take to get people hooked on using your company’s chatbot? For now, we’ve concluded that it has a lot to do with empathy and obsessive optimization.

[aditude-amp id="flyingcarpet" targeting='{"env":"staging","page_type":"article","post_id":2026231,"post_type":"guest","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"bots,","session":"C"}']

Our chatbot, Icon8, drew in hundreds of thousands of users in its launch week. Within seven days of launch, we had attracted 300,000 users by leaning on three main points: 1) The bot is all about selfies; 2) Virality is baked into it from top to bottom; and 3) Messaging apps have their own native sharing features built in. These three factors came together quite harmoniously, allowing our audience to expand rapidly. Here’s a sample video of the bot:

People use Icon8 because they want to enhance their selfies with our neural styling process. The bot uses a neural network to digitally “paint” any picture you send it, adding drama and gravitas to even the most run-of-the-mill photos. As easily as you can send a friend an image from your phone, you can rework your selfie into a picture deserving of a frame. It’s like adding an Instagram-style “filter” that intelligently enhances your picture and makes it look like a painting. You can choose from four to six different filters per day, and we rotate them daily from a pipeline of more than 100 filters.

AI Weekly

The must-read newsletter for AI and Big Data industry written by Khari Johnson, Kyle Wiggers, and Seth Colaner.

Included with VentureBeat Insider and VentureBeat VIP memberships.

Let’s unpack each of these three factors more deeply.

1. The bot is focused on selfies

Some 30 percent of all pictures taken by people between the ages of 18 and 24 are selfies. We spend an average of 54 hours a year taking them, and these pictures generally get the most likes, shares, and attention on social media. The human face holds an interesting power online — not only do eye-tracking studies reveal that faces are some of the most attention-grabbing images on the internet, but the sight of a face creates genuine human empathy. Selfies are nearly universal on the 21st-century internet, and our bot imbues them with a new layer of gotta-have-it novelty. Icon8 images get shared more often, further driving our growth.

2. The bots capabilities change every day

It’s self-evident, but neuroscience backs up the fact that people love experiencing something new — a neurotransmitter called dopamine rewards us every time we do something novel. This is why we update Icon8 literally every day. Though the bot only offers four to six different filters at a time, we rotate them and add new ones daily. People don’t like complex choices because of the increased cognitive load they require, so we don’t want to bombard the user with tens of filters to choose from. In a perfect world, the user only needs a few different filters that work perfectly every time, so this means that filters need to be personalized. We track who uses which filters and how often they do so, and we make decisions on which filters to add or change based on these metrics. This of course drives user retention and growth.

3. We baked virality directly into the bot

Starting from the hypothesis that people want to share things that make them joyful, we implemented a process for A/B testing our different neural styles. We tweak our neural network’s image-altering parameters as we think the users would most like, test the results internally, then release them into production to measure how the users respond. This means our users get the best possible iteration of a neural style, every time. Furthermore, messaging apps include fundamental sharing features that readily enable the spread of beloved content among their users. Some styles can look beautiful without being shared a lot, while others might be unusual, generating lots of shares and inbound users. We measure it all and optimize constantly. Our bot-generated images are watermarked, so people can easily find where an image came from if they love it — it makes discovery far easier than it might be otherwise.

If you want to test our chatbot to see what Icon8 can do for your selfies, you only need to send it a message here on Facebook or here on Telegram.

VentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Learn More