The Transform Technology Summits start October 13th with Low-Code/No Code: Enabling Enterprise Agility. Register now!
Last year in the lead up to F8, Facebook’s annual developer conference, the buzz was that Facebook Messenger would spark a “chatbot revolution.” Bots, some believed, would fundamentally change commerce and the nature of how businesses and customers speak to each other.
There was open speculation at that time whether Facebook Messenger and its 900 million monthly active users (now 1.2 billion) would embark on a journey to kill apps and achieve some sort of digital metamorphosis to become the next WeChat.
Well, it’s been a year of bots on Messenger, and while, to be fair, economic ecosystems aren’t built overnight, few of the most ambitious visions of bots on Messenger have come to pass.
There’s commerce on the Messenger platform, including advertising plans and the recent addition of group payments, but to support the creation of a bot economy for its platform, Facebook works with business partners and a community of tens of thousands of developers making Messenger bots.
Messenger developers have made ways to buy food, sell insurance, send plane tickets, save money or make global money transfers. Beyond business interests, there are also Messenger bots that help people register to vote, seek asylum in the European Union, and apply for U.S. visas.
Through the early hype and changes to the Messenger Platform, developers have worked to build what both Facebook and the developers hope is the killer bot that will go viral and drive mass adoption.
So ahead of F8, VentureBeat spoke to developers working under the Messenger Platform hood to hear what they think needs to happen for Messenger and bots to succeed and spread. Along the way we found developers whose bot outperform their app, met a startup CEO who blames poor developer outreach for lost sales, and virtually without exception heard people clamor for Facebook to give the world better ways to find and explore bots.
BotCompany: Improve discovery and developer relations
BotCompany is a startup with a team of six scattered across major cities in Italy that makes Telegram and Facebook Messenger bots for business customers. CEO and cofounder Emanuele Capparelli told VentureBeat that even though his company sometimes makes bots with natural language processors that can understand and exchange messages with users, he’s glad Facebook decided to take the chat out of chatbots. Last month, Facebook released Version 1.4 of the Messenger Platform. With the update, bot developers now have the option to hide the text input area and increase reliance on buttons, a menu, and a guided experience.
Bots that aren’t smart shouldn’t say a word, he said, and changes implemented in version 1.4 were necessary to make “mass market adoption possible.”
“I think a chat-free bot is better than a bot which pretends it can handle messages and then fails 99 percent of the time; plus it’s a needed move for Messenger bots to spread,” Capparelli said.
Removal of chat from the equation makes it easier for people to think of bots the way they think of an app or website, instead of as some all-knowing bot that can talk about anything or address every question. No talk means users will have expectations more in line with what a bot can actually do today.
“That’s why they [Facebook] are shutting down chat capabilities,” he said. “This [the option to type a message to a bot] is related to the quality of existing bots.”
The changes Facebook made with version 1.4 were necessary, Capparelli said, but he wishes the rollout of Messenger Platform update had gone differently. Capparelli said version 1.4 was issued with zero warning. Bugs emerged in the persistent menu of BotCompany’s bots, and as a result the company lost sales, he said.
“They completely changed the UI of bots without prior notice, and we lost a couple of sales because potential customers couldn’t use the bots as they immediately became unusable and ugly,” Capparelli said. “I don’t know Facebook’s policy with dev relations, but I personally think they are not doing the best they could. Nobody tells you anything before it’s too late.”
How to find good bots
Capparelli thinks two of the biggest areas for improvement when it comes to the young bot ecosystem are discovery and developer relations.
People may find apps to download from a recommendation by a friend or they see it in a news article, but a 2014 survey found that the majority of Android users got their last app from the Google Play store. There customer reviews, download rankings, and apps featured by the Android team help the cream of the crop to rise.
Ahead of last year’s F8 conference, there was speculation about a Facebook Messenger bot store. Message.io CEO Tom Hadfield described the rumored store as “arguably the most consequential event for the tech industry since Apple announced the App Store and iPhone SDK in March 2008.”
No bot store has meant that although tens of thousands of developers are currently making Facebook Messenger bots, there is no central place to find the highest rated or most popular bots.
Instead, in the first year of bots on Messenger, you could find bots with a direct link, a QR code, by search in Messenger, News Feed ads, or visiting a Facebook page.
Seth Rosenberg, a former product manager for Messenger, once told us that a bot store was never created because the ecosystem lacked “a wide range of great experiences,” and when a wider range of experiences became available, Messenger bots would be made discoverable “in contextual ways.”
A future version of Facebook Messenger, Capparelli said, should have some kind of area dedicated to bots discovery so users know where to find noteworthy bots.
“Look at bots. They’ve been there for a year and still aren’t shown anywhere in any of these apps, except some obscure references in the Messenger search bar,” he said. “If they put a top section for bots as they did with Stories, there would be swift mass adoption.”
The wrong kind of discovery
Discovery is also a big issue to Shane Gau. As the creator of Qwazou, a bot that helps people discover other bots, Gau often sifts through the recommendation area on Facebook Messenger, saying it’s the only place available on the platform for noteworthy bots are featured. To access this section, users have to tap on the search bar within Messenger.
While going through the bot recommendation area about a month ago, Gau began to see some sexually explicit bots serving up clickbait, and sometimes pornography. In the course of speaking with one of the recommended bots, we were sent to a wet T-shirt contest at a strip club that may have been outside Facebook Community Standards.
“Featured bots are surfaced by a combination of algorithms that detect the most highly engaged bots along with some curation from our team to surface new and interesting experiences,” Facebook said at the time in a statement shared with VentureBeat. “When a bot or its content is found to violate our Community Standards, it is immediately removed from the platform and in this case, the bot appeared as a result of the algorithm. We have reviewed our current featured bots and are in the process of/have removed any that are found to be in violation from our platform and will continue to monitor for any additional instances.”
In the days following the report, bots featured in the recommendation area dwindled, from the 20 bots that were normally there to one or two, or none at all.
It’s really not shocking that porn or clickbait would make its way onto a platform as popular as Messenger. What’s shocking, Gau said at the time, is what it appears to say about how much Facebook is concerned with bot discovery.
“There has been a lot of talk about how bots have never really met their hype. I am not sure how including pornographic bots in the bot recommendation drawer in Messenger is helping the bot cause for the rest of us bot developers,” he said.
Bots are better than apps and group chat
One of the biggest predictions about bots within the last year has been that bots would replace apps. Apps aren’t dead, but there are instances of bots outperforming apps.
That’s what happened to Swelly, a bot and app that let businesses or groups of friends create and share visual A/B opinion polls.
Swelly is made by Dvel, a marketing and tech company with offices in Santa Monica and Vienna, Austria.
Swelly has two apps and bots on five chat platforms but says 56 percent of active users comes from Facebook Messenger, Dvel CEO and cofounder Peter Buchroithner told VentureBeat in a phone interview, and he thinks that number can grow. Since launch, 2.7 million people have used Swelly.
“For us, I think Messenger is the biggest growth opportunity,” Buchroithner told VentureBeat in an interview. “They [Messenger] grew 300 million users last year. Messenger is huge, and I think that’s why for every demographic and geography you find users there.”
Swelly recently launched a group chat for its iOS app so people can ask friends which dress or pair of shoes to wear. Buchroithner is excited to hear that reportedly Messenger will soon allow bots to participate in group conversations.
The lack of a group chat option is something we heard early on as a gripe among bot developers. Sports Illustrated bots made by GameOn for the Olympics last summer saw highest levels of engagement on platforms that allow a bot to chat in groups, CEO Alex Beckman told VentureBeat. It’s also something the Microsoft bot team has found to be a factor in the success of Xiaoice, a bot launched in 2014 that still attracts tens of million of monthly active users on WeChat in China.
As Swelly explores ways business clients can use its bot, Buchroithner said he wants Facebook to provide access to more of its social intelligence so Swelly can target its visual A/B toward specific demographics.
Swelly has enjoyed growth on Messenger, Buchroithner said, because in addition to there being no need to download a bot, when you share the bot with a friend inside Messenger, one tap of the opinion poll or Swelly icon puts them in a chat window with Swelly bot in order to answer their friend’s question about which dress or pair of shoes to wear.
Buchroithner said he thinks of bots today more like a web browser than some sort of intelligent, advanced technology. Like Capparelli, he thinks bots should master being simple before they are able to deliver on the promise of doing more complicated things like the delivery of powerful AI or being great conversationalists.
“I think it’s a smart move by Messenger to make it a developer’s choice to disable the text layer to write a message because then users can realize a bot can really be anything from a lightweight app to kind of a website within chat,” Buchroithner said.
NearGroup: Share your bot
NearGroup is a Facebook Messenger bot that connects people who live near each other for chat dates. Photos are initially hidden, but after 15 minutes you can choose to see the other person’s picture.
“After 15 minutes of discussion they can disclose their identities and meet in real life, so it’s basically giving less focus on image and more on chemistry and talk,” NearGroup founder and CEO Prashant Pitti told VentureBeat in a phone interview. “The way all the dating apps are going on right now, it’s almost like body shopping. You swipe people left and right, judging them way quicker than you should, and we’re trying to change that by letting people decide on the basis of conversation rather than images.”
Mostly on Facebook Messenger and primarily in parts of Asia like Vietnam, India, and the Philippines, more than 200,000 people chatted with NearGroup last month. Its Messenger bot is largely responsible for the company’s early success, outperforming all other platforms including its Kik bot and Android app.
“We’re one of those use cases where the chatbot is working way better than the app,” he said. “We had our own app, and we were never able to get this kind of engagement. Bots are working very well for us.”
Unless you’re in a part of Asia where NearGroup is most popular, chat with the bot today and you will likely find no initial matches. But the bot can sit in the background until a person in your area shows up and NearGroup bot sends the user a message.
Ask a person if they want to chat once a match has been found and many say yes, Pitti said. With apps you never get the chance to ask, because after two months the app has more than likely been deleted.
“In the beginning getting the demand is very hard, to get both the guys and girls in one particular area. However, our chatbot framework allows us to basically wait for demand and supply to get at a good number,” he said.
The fact that bots have no download requirement also makes Messenger more accessible for users in places where smartphone penetration is low, connectivity is slow, or little cell data is available, Pitti said.
NearGroup has been able to grow audience with no advertising, Pitti said, because it requires users who want to speak with more than one person a day to share the bot with a friend.
Pitti thinks the smart thing to do is to make bots that don’t overpromise on an artificial intelligence dream of fluent conversation with a computer that has not yet been realized.
The challenges Facebook has encountered with M is evidence that getting AI-powered bots right is going to take time, Pitti says. The Information reported in February that about 70 percent of messages handled by Facebook’s M intelligent assistant required human intervention.
“When a company like Facebook is grappling with getting the AI work done, no wonder the chatbots focusing on the AI are having a tough time satisfying consumer needs,” he said.
The future of bots on Facebook
Ahead of F8, USA Today tech columnist Jefferson Graham summed up year one of bots on Messenger as disappointing. Messenger VP David Marcus admitted bots were overhyped and fell below initial expectations last fall, but in recent weeks renewed his commitment.
“We’re doubling down and tripling down on bots,” Marcus told USA Today. “We’re fully invested in this, and in it for the long haul.”
Facebook’s not alone in providing a bot platform — Google, Microsoft, and Amazon have also extended support — but it may be alone in being one of the only bot platforms that is unable to point to a single place and say this where you find amazing bots on Facebook Messenger.
Microsoft has its Bot Directory, Kik has its Bot Shop, there’s an iMessage App Store, and the Google Home and Alexa apps can give you a list of their actions and skills. Virtually every one of these offerings is less than a year old, and may have faults, but at least they exist.
It’s been suggested that the app store approach may not be the best fit for bots, that bots should not immediately get shoehorned into the same App Store approach taken with apps. Last fall General Catalyst partner Phil Libin said he believes bots will get famous like people, and as Swelly and NearGroup demonstrate, sharing bots with friends makes a difference.
Whether it includes Messenger Bot Store, becomes part of M suggested replies or by some other means, let’s hope that in year two the double or triple down includes a way to find the best bots available on Facebook Messenger.
While developers said it’s important to manage expectations surrounding bots and AI and applauded recent changes to the platform, the biggest complaint VentureBeat heard was that Messenger needs to fix its discoverability problem.
The social network isn’t just one of the earliest large tech companies to bring bots to its platform — it also owns the two largest chat apps in the world. In the minds of some people, as Messenger goes, so too does the bot industry. Its size and continuing evolution make the impact of any change to Messenger intrinsically affect the fate of a broader industry.
“The future of bots in the West is tied to Messenger. Of course there is Line, Viber, etc., but they really don’t count as a meaningful medium to bring this mainstream,” Capparelli said.