Conversational UX – The Digital Assistant for Intelligent Enterprise
Published on : Tuesday 01-06-2021
User experience for enterprise apps can be taken to new dimensions using conversational AI, says Suparna Deb.
Artificial Intelligence and Machine Learning is revolutionising the user experience of the digital world. Natural Language Processing is the arm of AI which deals with teaching computers how to understand and generate human language thereby making interactions with machines or software more ‘human-like’. This paper describes how NLP has been used in the context of enterprise software applications to simplify and smarten end user experience. It also covers how chatbots can integrate people and process in an enterprise via natural language. This bot integration of enterprise backends (like ERP, HR, CRM, Finance, etc.), with collaboration channels (like Slack, MS Teams, etc.), leads to evolution of the virtual digital assistant of an intelligent enterprise.
One of the keywords to success (or failure) of a product is Experience – be it customer experience, employee experience or user experience1. To beat the competition, heavy investments are being made in the field of user experience, thereby bringing in a paradigm shift in the software industry. From the days of end-users trying to understand, learn and respond to application behaviour it has come a long way. Now, it is the machine trying to understand, follow and respond to user behaviour. All this has been possible due to significant advancement done in the fields of Natural Language Processing and Machine Learning.
The results of this year’s McKinsey Global Survey on artificial intelligence (AI) suggest that organisations are using AI as a tool for generating value. These companies plan to invest even more in AI in response to the Covid-19 pandemic and its acceleration of all things digital. The whole objective is to enhance the productivity of employees and enhance the efficiency for serving customers. To increase the productivity of employees, they have to be enabled to perform their daily tasks in an efficient manner, find right information and act on that. Customers on the other hand expect a smooth interaction. 24/7 support can enhance the efficiency manifold. In order to generate value and enhance productivity it is imperative to transform into digital. Be it employees or customers, simplification of day-to-day tasks needs the efficacy of a digital assistant. Retail, HR, Commerce, Customer Sales and Services are few of the top sectors which are reaping huge benefits by moving to conversational user experience. There are over 100,000 chatbots just on Facebook Messenger alone. In March 2017, Microsoft announced that it has seen a 350% yearly increase of monthly active users to about 0.7 million users on its Cortana digital assistant2.
Following are the few value propositions that a digital assistant provides:
a. Enhance user satisfaction by automating time-consuming, tedious tasks from processes (HR, IT, procurement…)
b. Simplify access to information and deliver personalised human-like conversations at scale, 24x7, from anywhere, on any device
c. Improve overall productivity by focusing staff resources on added-value tasks, and
d. Scale businesses by reduce development efforts and costs within support teams.
So what are the typical characteristics of an enterprise digital assistant?
The digital assistant is intelligent. It enables the user to engage in smart conversations in plain natural language also known as utterances. It is the task of the NLP engine inside to parse the utterance and derive the business intent of the user using NLP. Once this is done, it invokes the App or Service mapped to that intent.
For example, an employee says, “I am not feeling well today and want to take a day off”. While parsing this utterance the NLP engine identifies want to take a day off as user intent. It maps this to Leave Request App/Service. It also identifies the values today and not feeling well to be matching the possible values for the input parameters Date and Type of leave. So, it decides to invoke Leave Request service with the current date and type it as sick leave.
This simplifies the entire user interaction and eliminates the need to learn how to use complex applications.
The digital assistant is single-point-of-contact for all needs of enterprise users. As explained above, it can invoke the appropriate Service while remaining independent across all applications. As a result, an obvious benefit for an enterprise is, the user has to now interface with only a single bot which in turn channelises or connects his requests to the appropriate enterprise app.
The digital assistant is context-aware. While engaging in smart conversations it remembers and understands the context in which the statement is being made. Accordingly, it frames a more accurate response. This part can also be implemented using rule based principles instead of machine learning. The NLP engine stores keywords from the conversations temporarily. When an utterance contains a reference word it browses through previous results and takes the absolute value for the reference. For example, when the user says, “show me products available”, the bot returns a list of products from the product catalogue. Next, the user says, “show me details of the second one”. At this point the NLP engine identifies ‘second’ as a reference word. Hence it picks up the second item in the list from the previous response and invokes the Read API to retrieve and show its details. This keeps the conversation close to typical human-to-human conversation. It also saves the user from repeating data or information.
The digital assistant is proactive. In the course of discussion, it can predict the user behaviour and thereby accelerate him in accomplishing his goals. Here two techniques are used. Most of the interactions follow a certain number of steps in a pre-defined order. This wizard-like behaviour is defined in the dialogue engine. There are several bot-builder platforms which help in modelling such dialogue flows. As a more advanced and dynamic technique, a deep learning algorithm is applied to learn and predict user behaviour to improvise on the initial defined flow. In the same product example illustrated before, digital assistants can be modelled to automatically propose follow-up actions on the read response, like creating a purchase order. It does not enforce rather guides the user proactively to speed up the entire transaction. As another example, bot learns from different usage of synonyms in utterances to map them to the same intent.
The digital assistant is omnichannel and consistent. Its behaviour can be integrated and made the same across all collaborative channels. The brain of the digital assistant is its NLP engine which runs on a central server. It can be deployed as a client application to different channels like browser based websites, Slack, Microsoft Teams, etc. The client bot app transfers the user request to the central NLP server. Owing to such architecture, the same conversational user experience can be provided to users of all channels at once. Also with this integration, the users do not have to leave their familiar work environment to interact with enterprise software applications. It eliminates waste generated via context-switching.
The digital assistant is efficient. “Bots have around a 12 per cent higher accuracy rate than humans because they are trained by call centre agents on actual user questions — and they don’t get distracted, tired or bored,” TechTarget stated recently3. As compared to humans, digital assistants can work round the clock and offer excellent returns especially in the customer sales and support scenario.
The digital assistant can continuously learn to improve its own performance over time. The task of building the bot does not end after putting in production. In reality the task to retrain the bot to learn from user behaviour and perform even better starts from there. It is imperative to build this feature into the digital assistant to keep it relevant and up-to-date.
It also hands over to human agents gracefully when it is unable to fulfil user needs. It is not practically possible for a digital assistant to know all the answers for its user. In such a case an elegant fall back mechanism needs to be implemented to handle the request. This can also include seamless handover to human channels like Intercom, etc.
The digital assistant is expected to work in an enterprise context and hence fulfil all enterprise qualities like security, etc.
The digital assistant has a persona. The persona is assigned in a way that it represents and echoes the brand value of the enterprise4. Extensive user research is conducted to study how users interact with enterprise software. It includes understanding their set of tasks, goals, pain points and needs. Next, personality models, like OCEAN five factor personality model5, are applied to define and test scripts with users. Similar research is also conducted to test voice pitch, range, speech rate and volume. This data is then used to determine and define a personality for a digital assistant in such a way that it reflects the brand of enterprise. For example, while conversing it is expected to be proactive, approachable, transparent, concise, respectful and using considerate language. It is very important for an enterprise to assign a persona to the assistant to protect its brand just the way it expects its employees to adhere to code of business conduct.
While AI is revolutionising digital transformation in many successful ways, it also brings out new issues about ethics and privacy. With the rise of machine power many dread that some of our most popular sci-fi movies like Matrix might become a reality. Two Internet giants just found out the hard way that chatbots learn offensive behaviour from users, as CNBC reported6. The software industry is now taking active measures to define principles that ensure digital ethics are followed in this era of rise of robots. Digital assistant is being designed with a personality model for a better future and with ethics in mind7.
Will the digital assistant or chatbots in general replace traditional apps?
The answer only lies in the future. While there is an ever-growing popularity and demand in conversational UI it will not replace the traditional GUI, at least, certainly not in the near future8. The main reason is that Artificial Intelligence has not yet arrived there. Deriving sense out of human conversations is not easy. Language is one of the most intellectual attributes of human being which distinguishes and differentiates him from the rest of the species. Over a period of time he has mastered the art to its current perfection. It is not easy to infuse machines with the same intelligence. Several iterations on improvisation of machine learning algorithms and NLP will be needed to bring the interaction to a satisfactory level. Till then, Apps would sometimes fulfil user needs in a much faster, reliable and simpler way with few clicks instead of going through elaborate conversations. If all the innovations that happened in the last five years are considered, it can be safely said that only time will tell which one survives in the long run.
User experience for enterprise apps can be taken to new dimensions using conversational AI. Instead of dealing with multiple apps with complex UIs, users can simply interact with one digital assistant which proactively guides the user to accomplish his business intents. This enhances overall productivity, efficiency and brand loyalty. While the concept sounds quite fascinating and futuristic, quite some work still needs to be done especially in the field of intent resolution and intuitiveness. More and more investment in the field of machine learning and NLP will eventually lead to the adoption and success of digital assistants. If this momentum in conversational user experience continues, the future of user experience is bound to be more humane.
Suparna Deb has industry experience of 16+ years in the software technology domain. Her last role was the Head of Product Management SAP Conversational AI, an AI based chatbot platform.
References/Bibliography
1. https://www.gartner.com/doc/3698955?ref=unauthreader&srcId=1-3478922254#a1641874572
2. https://blogs.gartner.com/adrian-lee/2017/10/10/conversational-artificial-intelligence-we-need-to-talk-about-it/
3. https://searchcrm.techtarget.com/feature/Customer-service-chatbots-help-reduce-product-returns
4. https://experience.sap.com/news/the-emotional-enterprise/
5. http://www.personalityresearch.org/bigfive.html
6. https://www.cnbc.com/2018/03/17/facebook-and-youtube-should-learn-from-microsoft-tay-racist-chatbot.html/
7. https://experience.sap.com/news/design-for-a-better-future-with-digital-ethics-in-mind/
8. https://chatbotsmagazine.com/chatbots-vs-apps-the-final-frontier-a0df10861c48