The chatbot industry was valued at about 17 billion USD in 2019 and is projected to reach over 100 billion USD by 2025 (Mordorintelligence, 2020). With a probable 34% cumulative annual growth rate, the industry shows promise in service automation. Besides, 65% of customers worldwide showed an inclination towards chat connectivity over customer call support (BotMyWork, 2020). However, the chatbot industry is facing significant challenges that are holding it back.
Chatbots have limited responses. Completely autonomous chatbots based on deep learning algorithms are yet to reach a scalable state. In many instances, chatbots require human input for addressing user queries. Usually, only a specific set of solutions are available via chatbots. Any issues beyond those require human intervention. The limited database can also throw users in a loop of repeating responses resulting in frustration. Chatbots are often incapable of addressing complex issues (Bluupin, 2020). Development and integration of advanced AI are cost and labour-intensive, which defeats part of the purpose.
Moreover, chatbots are not practical for all businesses. Some businesses are far too complex to integrate chatbots for serving customers feasibly. Firms specializing in luxury goods attract customers who expect personalised services for the premium they are paying. In such cases, chatbots give a bad rap to the associated brand. Only 13% of the customers worldwide expect to buy an expensive item through chatbots (Medium, 2020). It shows a clear indication of the reluctance of luxury item consumers to use chatbots for receiving services. These can also be cumbersome, as they have to zero in on probabilities based on customers’ chronological responses. Individuals requiring swift solutions are not keen on such services. Besides, 37% of customers worldwide expect chatbots to provide quick answers during emergencies. If users have to go through a series of questions before reaching the required information, then the outcomes are counter-productive.
Chatbots have proven successful in repetitive tasks rather than novel ones (CB Insights Research, 2020). Changing patterns of requests highlight their incapacity. Modern consumers can have a varying set of requirements, which cannot be addressed with specific types of requests that are chatbot-friendly.
Chatbots require constant maintenance. As a company grows, the range of interactions expected by associated stakeholders also changes. Hence, setting up a chatbot and forgetting about it is out of the question. Chatbots need to evolve with the related company, requiring constant monitoring, integration with updated databases, and accumulated information for tailored and insightful responses. The lack of feelings and emotions in these instant chatbots also inhibits them from addressing user distress with expected empathy.
Chatbots for learning purposes has several vulnerabilities. They can be tricked and bypassed to advance in associated learning programmes. Language learning applications utilise chatbots for receiving input and providing feedback on a learner’s progress. As the human element is missing, computer-generated responses that are more accurate than a new learner’s responses can trick these chatbots. Chatbots for education has yet to prove its feasibility as linear programs are incapable of addressing academic issues arising from various contexts. The mere exchange of information is not always adequate for practical educational experience. The shortcomings are mainly attributed to the reliance on “decision trees” rather than “advanced AI” in chatbots for education.
Chatbots were supposed to replace human beings as a mode of delivering services. However, the lack of human element has neutralised the conditions. Human agents need to have digital fluency and capability to work alongside chatbots to stay relevant.
Existing stereotypes inhibit the potential uses of chatbots. They are generally considered to apply to customer service sectors only (Medium, 2020). Conversational AI is being used across financial institutions and banks to deal with repetitive requests regularly. Even though they can have significant education and healthcare applications, their untapped potential is yet to be explored.
Early chatbots have often failed, as they were not able to process diverse responses. This gap has been minimised significantly, but the associated complexities have also increased. Modern chatbots are highly specialised but still operating within vocabulary and documentation constraints (BotMyWork, 2020). They are better suited for structured operations that function with limited information. Unstructured processes may be too farfetched and, in most cases, inapplicable for chatbots.
Users generally have a limited attention span. Hence, after a few minutes of interacting with chatbots, users either lose interest or are keen to speak to human representatives (Medium, 2020). Under such circumstances, creative measures for intriguing user interest plays a vital role.
Voice assistants are iterations of chatbots with a similar set of variables. Natural language processing capabilities limit them. Unless these assistants are maintained by large corporations like Google or Amazon, having a sustainable financial backup, technical expertise, and access to state of the art technology, they cannot reach their full potential.
Chatbots are smart platforms that can have huge upside when appropriately harnessed. But we are far from reaching its full fruition. A paradigm shift to advanced AI, which can harness human-like qualities and capabilities in addressing issues, is necessary. The channels associated with these chatbots need to be secured, and users should have the autonomy to place queries beyond language processing limitations (Mead, 2020). Conversational UI has the potential to be used as a tool to realise creative measures for addressing issues that significantly impact society, and to lead a technological disruption.