What makes a great chatbot?

For the purpose of this article, we will only consider computer-human relationships. Chatbot to Chatbot discussion is another hot (and controversial) topic especially with the recent Facebook AI shut down, whose AI program started to speak in its own language. For more information on this topic, I recommend the following link:  http://www.theepochtimes.com/n3/2274480-facebook-shut-down-ai-after-it-invented-its-own-language/.


But for now, let’s focus on the following issues:


What makes a chatbot successful? What features should a company look for when building a bot to attract new customers and keep its current ones?


– Obviously, a two-way communication channel is a great competitive edge for a business[1]. Each customer or potential customer, can interact with a “virtual representative” and get personalized feedback. Chatbots squeeze between the traditional one-way communication websites and contacting the company directly. Although current conversational AI programs are not as satisfying as human-to-human interactions, they can provide useful information to the user, collect precious data for the company and leverage it to ensure a better user experience [4] and, as it is an automated program, their response time is much faster.


– Lark’s chatbot [3] acts as an “AI nurse” and focuses on patients with chronic diseases like diabetes or heart failures. It not only provides useful medical advice but also creates a real human connection with the user. During a keynote presentation at Stanford, Julia Hu [2], CEO and Founder at Lark, mentioned what makes a great conversational AI. According to her, one can narrow down its research and focus on four main features:

– Contextual Data to create a History using mainly passive data to gain knowledge and make better decisions

– Good UX (user experience) as it improves the ease of use of the product and “humanize” the conversation.

– Personalization: One must leverage the data it collects by learning from past experiences to create a personalized and hopefully long-term relationship with the user.

– And finally, as obvious as it sounds, the content must be helpful and meet the needs of the consumers.


– So far, we only mentioned how humanized, a conversational AI can become. But it’s not just replicating the “human-touch” that makes bots successful. Their ability to trigger task is as impressive, if not more. Whether you need to call a nurse with Lark or to reserve a table at a restaurant, triggered task chatbots can actually do things for you. Advanced bots can implement speech recognition to prevent the user from even typing their requests and complex task interactions to define more precisely the user’s needs. Siri, Alexa or Cortana are the most popular ones as they managed to gather these techniques into one bot.




To conclude, a good chatbot makes your life easier on the consumer’s side and improve your business’s efficiency on the company’s side. I believe the next milestone regarding bots will be reached when conversational AI programs will be able to pass the Turing test which can be defined as an experiment that succeeds in simulating the “presence of mind, or thought, or intelligence in a machine. “[6]





[1]: https://medium.com/caravan-blog/are-chatbots-helpful-in-b2b-part-2-a-look-at-the-5-advantages-of-chatbots-f27467e1cc1e

[2]: https://www.linkedin.com/in/juliahuceo/

[3]: http://www.web.lark.com/

[4]: https://www.forrester.com/report/Pivot+To+PersonFirst+Personalization/-/E-RES135945

[5]: Source for the image: Deloitte University press

[6]: https://chatbotsmagazine.com/how-close-are-chatbots-to-pass-turing-test-33f27b18305e




4 comments on “What makes a great chatbot?”

  1. Hi Victor,
    Interesting article and thanks for sharing! Two thoughts came to my mind after reading your article:
    1. You mentioned AI chatbot in the health care industry, and I think this is indeed an industry that can really benefit from the use of chatbot. The key value proposition of chatbot here is the ability to give quick responses and provide authentic/trustful solutions, and this applies really well in cases of emergencies when people have to diagnose the symptoms accurately. I am from China originally, and we have a health care system different from that in the states, where hospitals are classified into sections by symptoms. People have to first choose the sections according to their own correct identification, and sometimes there are a lot of confusions. In this case, a chatbot can really be helpful.
    2. You mentioned several voice control assistants such as Siri/Alexa/Cortana, and this thought came to my mind which is really interesting. Nowadays most chatbots are based on a traditional GUI (graphic user interface), but if the technologies of natural language recognition can be developed to a more precise stage, would the shift from GUI to a complete voice-controlled interface be beneficial? In cases of building human connections, I think maybe the voice-based interaction is more helpful to make people feel warmer and friendlier. It is indeed an interesting topic to consider and an important decision to make between the two.

  2. Loved the picture, Victor! ‘Her’ is an amazing film.

    Since chatbots are becoming more complex and authentic, do you think our interaction with them can develop more? So far we’re using them predominantly for entertainment, and as well they are becoming increasingly popular within medical technology, psychology, etc. For what other aspects of our life would you predict a potential usage?


    Users who have LIKED this comment:

    • avatar
    1. Thanks for commenting.
      I honestly don’t know. People tend to think that conversational AI will revolutionize healthcare. I, personally, believe that entertainment will be the core business for chatbot companies. There’s a strong correlation between individualism and depression/loneliness. Bots might be a good alternative for that, and although depression is not a “hot topic”, it concerns a large section of the population.

    2. Actually I think it is happening now. Eugenia Kuyda from Replika came here to Stanford as a speaker for a seminar about chatbots hosted by Andrew Ng. She talked about how her company, Replika, focuses on developing an app that can be “a friend that is always there for you”. She sums this up as someone that talks to you, keeps a diary for you, helps you discover your personality. You can even pick their personality and the more you talk with it the more natural and friendlier conversations become. It involves nurturing your friendship and the app’s AI system. This idea came after one of their closest friends died in a car accident. They collected his texts and trained an AI that was able to talk like him. After developing the first version of their app, they started testing it with a group of users. By gathering data and vlogs from their first users, they found out that these users fell more comfortable talking with the chatbot about things they wouldn’t talk with any other human and helped them develop in areas that they hadn’t explore about themselves. I think another area that has potential is education. By automating certain tasks and giving more insights about students, teachers/tutors could improve the design of courses using ai by giving students a more personalized/individualized experience. Data generated by an immersive approach of education by using AR/VR systems for example could provide insights about how people learn the best, potentially giving educators, innovators and policy makers more ideas on how to transform the current education system.


Comments are closed.