What is Conversational AI?

By: Austin Bankston

So, what is conversational Artificial Intelligence (AI)? According to Georgian Partners, conversational AI refers to the use of messaging apps, speech-based assistants and chatbots to automate communication and create personalized customer experiences at scale (Brenier & Pria, 2019). Smart speakers, such as Alexa, Cortana, Siri, and Google Home, are stimulating the conversation surrounding the AI revolution. Conversational interactions with AI platforms such as smart speakers are driven by words, whether that be in a phrase or full sentences. Unlike social media, conversational platforms can support engaging, two-way interactions with private audiences.

Why is the future of conversational AI bright? All over the world, businesses are in a frenzy to unlock the potential of conversational AI, to lead to a boost in customer experiences, and in turn, loyalty. A recent report from Deloitte expects the conversational AI market to reach $15.7 billion by the end of 2024 (Comes & Murphy, 2019). This report reflects the high demand for technological innovation like conversational AI.

What will conversational AI look like for future generation? The evolution of virtual assistants, in thought and ideation is growing rapidly. When combined with automation and AI, these interactions can connect humans and machines through chatbots, smart speakers and virtual assistants. This integration of technology leads me to wonder: will computers be able to empathize with humans? I hope so! For example, The Avengers film portrayed an AI device, JARVIS, with human to machine interactions displaying conversational and emotional depth. Given its bright future, let’s explore its potential further.


The first chatbot was developed by MIT professor Joseph Weizenbaum in the 1960s. It was called ELIZA (Ina, 2019). The foundation of chatbots in AI has evolved since that time, including technology like Natural Language Processing (NLP), Machine Learning (ML) and AI. The evolutionary journey has developed into live onscreen pop-up chats. Digital assistants today help businesses scale AI to provide more convenient and effective interactions in business-to-consumer relationships. These interactions come directly from customers’ digital devices.

AI excels at automating simple and repetitive processes. However, some chatbots are still plagued by mediocre performance. One example of a bad chatbot system is healthcare websites. These healthcare chatbots struggle to interpret and provide detailed responses to users due to complicated medical terminology and the vast number of variables to consider. The demand for more precise responses exceeds the chatbot’s capabilities. This is bad for both the business and the customer. For AI to be efficient, the chatbot has to streamline its decision-making process.  So, how do companies increase the efficiency of chatbots? Well, the answer may be simple. Developers should be able to find a work-around to the limitations of having a vast number of variables within its operating system. Could it be as simple as an additional route for the users, such as adding a live agent? The live agent could suggest a different question or issue.

As AI technology and implementation continues to evolve, chatbots and digital assistants will become more seamlessly integrated into our everyday experiences. All chatbots use data, which is accessed from a variety of sources. If the chatbot is developed correctly and the data is of the highest quality, this will make the chatbot efficient. On the other side of the coin, if the data is not the best quality, there will be limitations to overall chatbot functionality. Chatbots that function well require machine learning models that are exhaustive in scope. From personal experience, I find the inaccurate or irrelevant information provided from chatbots incredibly annoying. So, at this point I would rather just deal with a live agent or call in to get the information that I need.

Smart Speakers/Virtual Personal Assistants

Let’s look back at the beginning of speech recognition in smart devices to improve our understanding of advancements this far. The technology that is being used today has been in the making for a while.  In fact, it has been more than half a century since the first voice activated calculator was invented. Before our smart speakers and phones could understand our requests to read the weather report, turn off lights in homes, play music, etc., machines had to learn how to hear, recognize, and process human speech. Advances in AI have come a long way since the earliest listening and recording devices.

Smart devices are also known as Virtual Personal Assistants (VPA). An IBM engineer introduced the Shoebox at the 1962 World’s Fair in Seattle (Jelinek, 2012). Shoebox was a voice activated calculator that understood 10 digits and 6 control words: plus, minus, total, subtotal, false, and off (Jelinek, 2012). Once the Shoebox was given instructions, it would calculate and print out the answers to simple math problems. The device attempted to recognize and act on the specific frequency of the vowels in each spoken digit. 

Today, one of the more prominent listening and recording devices is a supercomputer, Watson. Watson was designed to parse massive datasets and return actionable information based on its understanding of natural language, combining technology from IBM and Google (Ferrucci, 2011). It was incredible to witness the capability of Watson answering natural language while on Jeopardy. I remember Watson taking on the grand champion, Ken Jennings, in an incredible trivia match. This competition, which Watson won, was a major step forward for both the speech recognition and software intelligence at IBM and Google. Watson’s creation was pivotal, as it sparked the development of Google Assistant with a focus on the consumer aspects of voice. Given radical advancements from these technological pillars, I chose to have a smart speaker in my home.

In order to understand commands received from users and provide the correct or desired response, smart speakers must effectively use natural language. A massive challenge companies face is programming these smart speakers to understand dialectal variants, accents, and slang across the world. I’ve occasionally asked Alexa a simple question and have received an answer totally unrelated to what was originally asked. Identifying the correct use of vocabulary is a critical part of making smart speakers successful. Another challenge that smart speakers have is understanding context, references, and previous data apart from natural language.

The next era of virtual intelligent assistants will allow people to have in-depth conversations with these machines creating human-like relationships. This is exciting for those of us who dream of having R2D2 from Star Wars or JARVIS from the Avengers at our beck and call for conversations! These conversations will include emotional depth, adding the element of advice and providing a personalized experience for each customer. Just as in the movies, the AIs, like JARVIS, will continuously learn users’ preferences on everyday use/lifestyle and adjust accordingly. Voice recognition and basic gesture recognition is already in practical use for several VPAs.

AI cannot be programmed with logic. Instead, conversational AI is trained through the principle of “trial and error”. The next advancement in AI needs to include better post-data processing techniques in order to learn, remember and predict the preferences of their individual users. The learning process of AI can be compared to that of a toddler. If a child touches a hot stove, it will learn to have cautious behavior towards stoves in the future. Therefore, only with experience does a child learn to think logically before acting. Just like a toddler learning from its experience, AI must absorb its users’ preferences. This entails learning the system users’ likes and dislikes to program reactions accordingly. In order to achieve this, the AI needs to have enormous data storage capabilities and the optimal algorithms to learn from its experience.

For user experience to be at its peak, the post-processing feature in AI requires the most work as it continuously accumulates knowledge. The next step for AI to be faster and more adaptive is having an intelligent assistant that predicts future actions of its user based on accumulated data. Personally, I am extremely excited to see this technology evolve and become commercialized. It will allow the human race to be much more efficient in everyday life.

Side Effects of AI

Although AI offers many advantages, there are still disadvantages that may pose barriers to the future development and adoption of conversational AI technology. One example is the fear of AI systems. No matter what, people will argue that the advancements of AI will destroy far more jobs than they can produce. This is a logical fear, but it is highly improbable. In addition, people will always fear the technological repercussions on humankind, such as technology taking over the world and wiping out the human race. Everyone has seen The Terminator or iRobot and understands the dystopian possibilities that the future may hold.

Additionally, the use of AI in surveillance technology, such as facial recognition in public places, is another problematic area. People are concerned about what personal data is being gathered and what it is used for. The technology has advanced so rapidly that the law and governments have failed to keep up. AI is superior to human intelligence, when large amounts of data and many factors such as processing power, machine learning programs, system controller, etc., come together. However, this should not cause concern because only humans can create and oversee the use of AI devices. Additionally, human users have the capability to think logically and distinguish between useful and worthless advice from their AI. With that said, these concerns do not present significant risk when compared to the systems benefits.  

What’s Next?

Recently, there has been a surge of conversational AI studies. Numerous companies from a variety of industries are pursuing conversational AI technology, resulting in rapid research and development in the field. AI learning is unlimited and limited at the same time. Machine thinking is limited by computer processing power. The main issue is that the logic and sense behind choices of what to do with processing patterns can only be recognized by humans. However, AI is unlimited in a sense because machine learning is vastly superior to the learning capabilities of the human brain. As machines continue to have more and more processing power, machine thinking may provide humans information that is too complex and would take an incredible amount of time to understand.

The next question is how can conversational AI be used to help data scientists with their responsibilities? This technology can provide data scientists with time efficient, educated and precise decision-making. For example, data scientists could implement AI systems for extracting, cleaning and analyzing gathered data. Similar to everyday users’ experience with using conversational AI to maximize their daily efficiency, data scientists can use this technology to make informed decisions and empower them with relevant knowledge.

Overall, the potential advancements in conversational AI will be incredible to witness and experience. After observing the transition from chatbots to smart devices, I know the possibilities are truly endless, and I look forward to watching future innovations unfold.


  1. Brenier, J., & Prial, J. (2019, May 16). An Overview of Conversational AI. Retrieved March 2, 2020, from
  2. Comes, S., & Murphy, T. (2019, September 20). Conversation starters. Retrieved February 15, 2020, from
  3. Ferrucci, D. (2011). A Computer Called Watson. Retrieved March 3, 2020, from
  4. Goasduff, L. (2019, September 12). Top Trends on the Gartner Hype Cycle for Artificial Intelligence, 2019. Retrieved March 1, 2020, from
  5. Ina. (2019, March 12). The History of Chatbots – From ELIZA to Alexa. Retrieved March 3, 2020, from
  6. Jelinek, F. (2012). Pioneering Speech Recognition. Retrieved March 2, 2020, from

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s