Chatbots Landscape At A Glance
Are chatbots overhyped? Let us take a deep look into the chatbot world…
Before we dive into the chatbot landscape, I would like to spend some time to discuss the issue that has been widely debated — “Are chatbots overhyped?”
Are Chatbots Overhyped?
Right after Tay.ai, Zo.ai, Microsoft launched “Ruuh AI“ this February on messenger, a chatbot for entertainment purposes. Just like every company in the chatbot industry, Microsoft is kept experimenting the chatbots capabilities in helping people and organizations achieve more through new conversation models. Over the past years, we can see more and more companies across industries — brands, e-commerce, financial services, airlines, and even government (gov.sg) — started to embrace the new conversational interface to engage with their users on messaging platforms.
While some believe that chatbots will reshape the messaging apps and will replace the current Apps, opponents argue that it is a waste of time and it’s an overhyped industry. I have a relatively neutral viewpoint, I believe chabots will reach significant growth in the coming years but the industry is still facing some challenges.
3 reasons on why chatbots will rise:
1. Chat is there: Messaging APPs are skyrocketing
According to Flurry’s report, users time spent in social and messaging apps grew by 394% over the last year (2016), which is the biggest driver of the 69% growth in total mobile app time spent. As messaging apps have become indispensable parts of our lives, enterprises are determined to be where their customers are.
2. Post-APP Era: We’re moving towards a new interaction model
Gartner research shows that users from the US, UK and China used just 6 to 10 apps each month in 2016. This represents a 6% year-on-year decrease and has led Gartner to declare the “post-app era.” Gartner also reported that Facebook Messenger and WhatsApp remain the most used mobile messaging apps with 81% and 61% respectively. In China, WeChat (95 %) maintain their leading position. Messanging apps have become incredibly “sticky” with users with 72% of mobile messenger survey respondents using the app at least once per day.
3. Still a key focus for industry giants
Industry giants are encouraging developers to work on their chatbot platforms. In April 2016, Facebook and Line both allowed developers and businesses to build chatbots for their messaging service. One year later, Facebook is launching the group chatbot service on the coming F8 (Note that Wechat, Slack, and Line already support group chatbots). Google launched the Google Allo last September; and Slack has a $80M fund for bots that run on its platform. From the above activities we can conclude that chatbots is still a key development focus for all the big tech companies.
The main challenge: AI technology is not there yet.
People who have experiences chatting with Siri, Allo, and Alexa must have the same question: Why they can’t understand my questions / sentiment and answer correctly? Why the chatbots can’t act smarter? The key reason is the natural language understanding (NLU) technology is not mature enough. Currently, some of the NLP issues like named entity recognition, part-of-speech tagging, and information extraction have been solved, some are in good progress, but some NLU issues like summarization, paraphrase, questions answering and dialogue, are still really hard.
Before we see major breakthrough with the NLU technology, we should let the users know that they are chatting with a chatbot that was designed to solve particular problems instead of general ones. As long as the customers are aware of it, they won’t expect the bots to answer everything and to understand their sentiment.
I believe it is more important to build a user friendly chat experience and interface than to make a super smart bot. After all, a bot that do one thing well is more helpful than a bot that do many things poorly.
Chatbots Landscape At A Glance
Here we finally cut into the subject — chatbots landscape. There are a lot of developed chatbots, bot frameworks and AI tools around the world. I have segmented the chatbot market into 5 main segments — Instant Messaging Platform, Chatbot Builder, IM integration, AI Tools, and Chatbot Applications.
Instant Messaging Platform
The rise of Chatbot is closely related to the change of user behavior. According to Nielsen’s report, consumers are spending over 85 percent of their time on their smartphones using native applications, but the majority of their time is spent using just five non-native apps. And the top five include social media, gaming, and instant messaging. According to Facebook, there are more than 1 million MAU on messenger in July 2016, people use it to connect with the people and businesses they care most about, and whatsapp passed over 1 billion MAU since Jan 2016, skype has over 300 million MAU, telegram has more than 100 million MAU, Line and Wechat also have 220 million and 760 million MAU respectively.
Chatbot builder is a platform that let users build their own customized chatbot through it without coding knowledge. There are dozens of chatbot builders globally and some of these startups already got funded, Kasisto, for example, already got $11.45M from banks and Mastercard. Some of the commonly-seen chatbots were built from these chatbot builders. CNN’s chatbot is built with Chatfuel’s platform; MasterCard partnered with Kasisto to build its chatbot; Octane build the Maroon 5 bot, and Compose.ai built the chatbot for an e-commerce company parenting selection.
2. Motion AI — has been used to create more than 9,000 bots and to send or receive more than three million messages, raised $700k Seed from 5 investors on December 16, 2015.
3. Octane AI — build bots on Messenger for Maroon 5, LindsayLohan and others; raised $1.5M Seed from 6 investors on November 2, 2016.
4. Converse AI — integrated with slacks, Messenger, SMS, Twitter and other messeging platforms.
5. Compose AI — Specializes in Asian languages understanding, building conversational bots for Asian EC companies, financial institutions and brands on Messengers, Line, Wechat and Viber.
6. Recast AI — a collaborative online platform that enables users to develop and share bots with others; raised $1.12M Seed on June 24, 2016.
8. Manychat -more than 170,000 bots have been created using the ManyChat platform; raised $125k Seed from 500 startup on December 27, 2015.
9. Botsify — building bots for brands on facebook messengers.
10. Flowxo — a bot platform for Messanger, Slacks, SMS, Telegram, and Web.
11. Howdy — a Slack bot that collects answers on your behalf and delivers them in a convenient report; raised 1.5+ million in October and December 2015 from Slacks and other investors.
12. MMUZE — shopping assistants for businesses on Messenger; raised $1.1M Seed on February 22, 2015.
13. Kasisto — a conversational AI platform powering virtual assistants and smart bots across mobile, messaging, and wearable; raised $11.45M in 2 Rounds in August 2014 and January 2017, backed by DBS, Mastercard and Wlls Fargo.
14. Meya.ai — build bots on Messenger, Twitter, Slack, Intercom, Kik, Telegram, Twilio and more.
15. Kore.ai — a bot-based messaging platform that brings individuals, teams, and business systems together in a single, streamlined interface.
Some of the chatbot builders support only one platform (Most of the time, it’s Facebook Messenger) while an IM integrated platform supports multiple instant messaging platforms and the user only need to build once to deploy the chatbots on different platforms. For example, Compose.ai helps users to build their bots on Facebook Messenger, Line, Wechat and Viber , Flow XO supports Facebook Messenger, Slack, Telegram and Twilio SMS while Converse.ai is integrated with Facebook Messenger, Slack, Telegram, and Twitter.
AI Tools (NLP / Machine Learning)
There are some existing NLP and machine learning tools for chatbot developers to train there chatbot. Giant tech companies Google, Microsoft, Facebook, IBM all have their AI tools. Besides, wit.ai (acquired by Facebook in January 2016) and API.ai ( acquired by Google in September 2016) are both great NLP training tools.
1. Wit.ai is a natural language tool for developers which claims that it supports 50 languages and is used by over 65,000 developers. Wit.ai suggests its users to train the bot with stories that describe the most probable conversations paths. Below is a training process example from its official website.
2. API.ai supports 15 languages and is integrated with messaging, virtual assistant and IoT platforms. Below is an intent determines example for pizza delivery, which is designed to understand some basic requests from users when they want to start their order.
As chatbot becomes an important communication platform for business, government and publicans to communicate with their customers/ audience, we can see more and more chatbots on different platforms, according to Facebook, there are more than 30,000 chatbots on Messenger (as of Sept, 2016). There are different types of chatbot applications:
- E-commerce: Amazon, eBay, Walmart, Macy’s
- Media: CNN, BBC, Bloomberg, Techcrunch, CNBS, BI
- Financial institutions: Wells Fargo, America Express, RBS, Bank of America, Capital One
- Travel platform: skyscanner, expedia, KLM, Lyft, Uber, Hotels.com, Kayak
- Brands: Burberry, Tesla, H&M, Tommy Hilfiger, Unilever
- Publicans: Lindsay Lohan, Maroon 5, Elon musk
- Government: Singapore government
1. There are 3 reasons why chatbots will rise: “Chat is there”, “The coming Post-App Era”, and “key focus for industry giants”
2. The main challenge that chatbots faced: AI technology is not there yet.
3. There are 5 key segments in the chatbot landscape – Instant Messaging Platform, Chatbot Builder, IM Integration, AI Tools, and Chatbot Applications
Feel free to let me know if any comments on the article. And if there are any great companies that I missed, please contact me through messenger / email and I will update the landscape accordingly. :)