Over the past year, bots have been the darling of Silicon Valley because they represent how quickly technology is advancing and the promise of artificial intelligence. But halfway across the world, there’s one bot focused on a very human notion: faith.
Azkarbot, which delivers prayers and offers daily prayer reminders, is one of the fastest growing bots on the Messenger Platform. Since its launch, Azkarbot attracted more than 15 million users in 4 short months. While the media discusses all the ways bots can benefit a business, Saif Ahmed, the creator of Azkarbot, is bringing together two often opposing concepts—technology and religion—to demonstrate how bots can make a difference with people.
Among the unique and varying pursuits of people in the Middle East and North Africa (MENA), one shared interest is the importance of prayer. Predominantly Islam, the region is home to approximately 315 million Muslims who pray five times—from sunrise to sunset—each day. As a 21 year old student at Helwan University in Egypt, Ahmed needed help scheduling these daily rituals between his foreign trade studies, his pursuit of a masters degree in economic theories, and his freelance work as a full stack developer.
First, he tried setting up calendar reminders, but its manual nature was an instant barrier because of the upfront effort required. Then, he tried a mobile app, but the notifications felt more bossy than utilitarian, and were easy to ignore. With prayer being such a personal practice, Ahmed wanted reminders that also felt personal, like they were coming from a friend.
Building on Community
Thinking that he wasn’t the only one who needed help, Ahmed decided to create a framework of his own, on a platform where people were already talking to their friends: Messenger. Before he started building his bot, Ahmed pretended he was a bot, and interacted with people directly. He started simply at first, using timezones to determine when to send morning and evening prayer reminders.
On October 1, 2016, Ahmed officially started developing Azkarbot and used the results of those conversations to inform and define the bot’s structure. Relying on Chatfuel to build the bot, five days were spent outlining the its requirements, specs, and scripts; another five were spent conducting more user testing, this time through the bot itself. During that time, Ahmed also shifted from full stack development to machine learning via Wit.ai, which enabled him to create small samples using natural language processing.
While he initially wanted his audience to be any Muslim who understands Arabic, Ahmed soon realized how many people in the MENA region were engaging with Azkarbot. They were interacting with it and offering feedback on a daily basis, all of which gave Ahmed insight into what people really wanted and needed.
Evolving a Prayer Bot
Not only were people using Azkarbot, they were also sharing it with their friends and posting about it publicly. To help it grow further, Ahmed visited 24 different cities in Egypt to speak about Azkarbot and the importance of chatbots at various student gatherings, companies and non-governmental organizations. He also gave social influencers early access to the bot and created incentives for people to spread the #Azkarbot hashtag, which reached more than 4 million people on Facebook and Instagram.
As awareness grew, so did user input and developer insight. Ahmed could see how he was solving a seemingly small problem for a large number of people, and how they preferred to interact with the bot. Behaviorally, people liked tapping buttons and quick replies more than typing. People also only used text input for simple words that described their feelings, which they also expressed via emoji.
It was those emotions that inspired Ahmed to add more features to Azkarbot: delivering prayers based on people’s moods and offering audio files, giving people the option to listen to a prayer instead of reading one. By watching how people were actually using and reacting to the bot, Ahmed discovered new ways to restructure the conversational UI and introduce new capabilities without changing the core functionality of the bot.
“You can’t manage what you can’t measure,” said Ahmed. “If Azkarbot remembers what people were feeling over the last few days, it can make suggestions to help with those feelings. It’s all about the dialogue—if content is king, then the conversation is emperor.”
Sharing Belief Systems
As Azkarbot became more successful, Ahmed started looking at other automated interactions. Using the same community approach that helped him, he created lists of other bots and asked people to use and provide feedback for each. Soon, Ahmed was helping nearly 95% of the bots in MENA enhance their experiences.
Today, Ahmed is consulting on more than 26 bots and finishing his final semester in school. When he graduates, he plans to focus on the company he’s also building: SocialHex, a platform that can help people build and manage bots of their own across multiple channels. When he needs to find somewhere to do all this work, he can use Co-Bot, another bot he’s developing that helps people find and reserve co-working spaces based on their location.
While Ahmed is mastering how to build automated interactions, he hasn’t forgotten the most important element in developing a bot: humans. By simplifying tasks that are meaningful and doing so in way that feels personal, Azkarbot has become a bot people can believe in.
Elena Ontiveros is a content strategist at Facebook, where she helps the Messenger brand sound and feel more human. Elena also defines best practices for businesses who build bots and shares her knowledge at conferences, including Midwest UX, Interaction 17, and Humanity.AI.