What is a Conversational UI?
A Conversational User Interface (UI) is a computer program designed to simulate a conversation with human users. This can be done over TEXT using a chatbot, also called chatterbot, or it can be done over VOICE, using a voice assistant. Some popular voice assistant examples are Siri and Alexa. They are increasingly used for business purposes in every industry. Conversational UIs are used in various industries to assist humans in their task to the point that we also name them virtual assistants.
Projected increased use of chatbots in 2021 and beyond
Conversational UIs are performing better and better and in 2021, the rise of AI powered chatbots continues. Chatbots are becoming increasingly useful and important in everyday business life. They help to reduce cost, save time and increase revenue. According to Emergen Research, the conversational UI market size reached 4.91B USD in 2020 and does not show any sign of slowing down. Believe it or not, it is expected to register a Compound Annual Growth Rate (CAGR) of 22.6%, during the forecast period.
Conversational UIs are used in an increasing number of businesses for various purposes. They can be used to schedule appointments or meetings as well as sending reminders. They can be your listening ear when you feel down, thanks to progress made in the field of Natural Language Processing technologies (NLP) or for customer service purposes. They also can be your assistant while you are shopping online and help you find the best deals. They are also used in marketing and to help increase conversion rate.
Type of chatbots and what powers them
Conversational UI can be rule-based or AI powered, depending on the features your business needs.
A rule-based conversational UI will operate using a decision tree. They are the most limited one. It does not mean that they cannot handle complicated tasks but the difference with AI bots is that they cannot operate outside of defined rules. In such a system, the bot will ask you questions and will narrow down the query by asking you more and more precise questions to give you a final answer. For example, if you are in the health business, you may benefit from a symptom checker to help you perform triage based on the patient’s state. In such a system, the bot may ask you where your pain is, then it could ask the patient to rank the pain from 1 to 10. Next question could be asking if there are other symptoms (eg. bleeding or fever, etc). As the user answers the different questions, the bot will be able to provide an accurate answer and set the patient priority to high, medium or low. Maybe the patient would need to go to the emergency room immediately or simply book an appointment within the next few days. Such a feature would be of great help to reassure patients and help improve medical services efficiency.
An AI powered conversational UI will have the power to improve itself over time using the data fed to him every time a user adds inputs to the system. This will improve the accuracy of the system. The way to add intelligence to a chatbot is the use of Natural Language Processing technology. This is a sub-domain of artificial intelligence at the crossroads with linguistics, statistics, computer science and information engineering.
Depending on the project scope, we can also include predictive models to solve various questions to which the Conversational UI would be a wrapper used to collect data and feed a model to solve a pre-defined problem. For example, in a medical setting, the chatbot could be used to collect data based on the answers given by the patient as well as some medical values to run a prediction on a Machine Learning or Deep Learning model and perform a medical diagnosis. This kind of software would be useful to know whether a person is in needs of immediate care or to improve post-surgery care, for example. Another use case for a Conversational UI could be providing guidance during rehabilitation.
Voice assistants will also need the use of Speech recognition and Speech-to-Text technologies in conjunction with Natural Language Processing technology in order to become ‘smarter’. Smart is between brackets here as true human-like intelligence does not exist at the moment in the field of Artificial Intelligence.
Artificial Intelligence is currently made of Machine Learning (a set of statistical techniques for problem solving) and Deep Learning. Deep Learning involves running data through a set of neurons and is in fact a subfield of Machine Learning. Deep Learning represents the closest imitation of how a human brain works with neurons. Machine Learning and Deep Learning are subfields of artificial intelligence which can be used to perform Natural Language Processing tasks. Natural Language Processing is also used for sentiment analysis tasks.
The computer program will develop a way to answer query after analyzing words and contextualize them together in order to make sense and to be able to answer customer’s queries appropriately. This is also the way Google search engine works. The search engine will look at a set of keywords in an article and will associate them together to make sense of the article by contextualization. For both the chatbot and the search engine, the use of Natural Language Processing will improve the overall user experience by handling the user’s request with a greater accuracy.
In the end, it is not a question to know whether you should use an AI chatbot or a rule-based one, you may be well off with a “simple” rule-based chatbot following a strict set of decisions. In fact, it may be a better use than an AI chatbot depending on your business needs. More complex and expensive does not necessarily mean this is the best answer to your business needs. Also, you do not necessarily need to switch to a full automated system from day one, sometimes a combination of a chatbot and live chat can be very effective and will provide an improved business workflow.
How can chatbots be improved over time
Bots initially start out fairly basic but the good thing is that it can be temporary and that we can improve the system with proper testing and data collection.
In this section, let’s use the example of a medical clinic. Of course a chatbot can be applied to most business models. This busy medical practice is trialling a chatbot for triage purposes in order to see the most urgent cases first and provide reassurance for patients that are only coming for a review. The bot can then potentially schedule them in for a telehealth consultation where you can review them at a less busy time over a video call or schedule them in for an appointment on another day, or it can alert the Dr in charge and schedule in a same day appointment if deemed urgent.
BETA bot release
After being tested, bots should be used on a small pool of staff and select patients who would be warned that they are in beta testing. Thereafter, after testing the bot under various circumstances, the bot would be released live. Steady improvement of the system is made by continually feeding the bot with new acquired data, helping to improve accuracy over time. Nonethelesee, the bot will have to be monitored regularly to prevent data and model difting. In such a case, the model loses accuracy over time. This can be due to many factors such as change of environment, change or skewness of data. This process can initially seem tedious and time-consuming, making it slower to implement effective AI Conversational UI but eventually the process will pay off.
Refine the algorithm
As the algorithm is refined using new data input of the type of patients and the most common requests, and because Natural Language Processing technology evolves daily at great speed, the bot will gradually evolve and grow to serve the specific and unique needs of your clinic. It will function as a 24 hour system to provide information, provide reassurance and alert the nurse or doctor should the need arise.
Regulations are important and necessary. Governments and medical regulation bodies are stepping in more aggressively in order to provide a safe framework to protect patient data. For example, it is possible to auto-destruct conversational datas after its use. However, the conversation can also be downloaded into the patient’s clinic records as part of medico-legal requirements. This can be done by the bot system without need for additional work time on the part of the doctor/clinic staff. This will also lessen the possibility of scribing mistakes.
Conversational UI, chatbots and voice assistants development are an exciting software solution with the potential to bring positive impact to any business right away. If you need help with the implementation of a conversational UI for your business, at Digital Nuage, we would be happy to assist you with a customized solution to your needs.