Review articleConversational agents in business: A systematic literature review and future research directions
Introduction
Technological investments in the field of business have shown an increasing interest in exploring conversational agents [1]. The estimates indicate that the application of these agents will be responsible for cost savings of $11.5 billion by 2023 [2]. These gains originate from the highly available, scalable, and cost-effective services provided by conversational agents [3], focusing on the better usage of technology [4] and improving service quality and market competitiveness [5]. Furthermore, these services increase user engagement by offering customized flexibility and experiences that are simple and natural [6]. However, like any other dialogue system, limitations arise from the grammatical complexity or semantics of the conversation, such as inappropriate responses generation [7].
In recent years, advances in artificial intelligence methods enabled an upsurge of powerful technologies, for example, deep neural networks. Hence, several works introduced these techniques in the dialogue pipeline, usually composed by Natural Language Understanding (NLU), Dialogue State Tracking (DST) and Natural Language Generation (NLG) [8]. Additionally, end-to-end frameworks became popular in both non-task-oriented and task-oriented dialogue systems [9]. Such methods, in combination with strategies of digital transformation, contributed to current research perspectives [10].
In addition, the state-of-the-art encompasses a more specific set of topics, for instance, self-learning and personalization methods. In this regard, some recent reviews focused on conversational agents from a technical perspective [9], [11], [12], [13]. At the same time, emerged business opportunities require more capable conversational agents (e.g., provided with human-like features) which have been considered by others reviews [14], [15], [16]. However, in the best of our knowledge, no literature review outlines the conversational agents in business domains, primarily taking into account machine learning methods.
This article addresses this gap with a systematic literature review that explores the following question: what is the role of artificial intelligence in the last decade on the state-of-the-art of conversational agents applied to business domains? Therefore, we conducted a literature review with a well-defined methodology to answer three general, three focused, and two statistical research questions. The contributions of this literature review are: (i) provide a mapping of the business domains explored since last decade, describing computational methods for self-learning, personalization and response generation of conversational agents; (ii) outline the state-of-the-art focusing on the business support through conversational agents; and, (iii) identify opportunities and future research directions towards more capable conversational agents in business.
The remainder of this article is organized as follows. Section 2 describes the related works. Section 3 details the materials and methods used in this literature review. Section 4 presents the results with a focus on the research questions. Section 5 discusses the limitations of this research. Section 6 approaches a discussion of the findings, followed by Section 7, which presents the opportunities and future research directions over the field. Finally, Section 8 provides the final considerations.
Section snippets
Related work
The aim of this study is to review the literature of conversational agents in the business domains with a focus on machine learning. In this regard, some surveys have been conducted to review conversational agents from a technical perspective.
The study proposed by Hussain et al. [11] focuses on chatbot classification and chatbot design techniques, where the authors explored task-oriented and non-task oriented categories of chatbot. These categories were taken into account in a discussion on how
Materials and methods
This section presents the methodology used to conduct the study. It follows the principles of systematic reviews, as proposed by Petersen et al. [17] and updated by Petersen, Vakkalanka, and Kuzniarz [18]. The methodology encompasses the planning and execution of the literature review, and employs five well-defined steps:
- 1.
Definition of research questions
- 2.
Definition of search processes
- 3.
Definition of the criteria for article selection
- 4.
Execution of data extraction
- 5.
Execution of analysis and
Results
This section presents the results of the literature review, structured according to the research questions. As suggested by Dalmina et al. [77] and Dias et al. [78], the previously described systematic review methodology helps mitigate possible biases in the results. First, this section outlines the methodology execution, associating each step with the number of selected articles. Then, it describes the main findings by answering each of the research questions.
Fig. 1 illustrates the filtering
Limitations
As any other literature review, this study presents limitations that may affect the scope of the results. However, the decisions taken during the planning and execution try to mitigate them.
In the planning of this review, six research databases were selected to reduce bias. These sources contain peer-reviewed publications from Computer Science and Information Technology. We manually defined the search string, considering two major terms and search terms. The search terms were selected
Discussion
This section presents a discussion of the findings based on the analysis of the results. The findings outline the literature of the last decade and illustrate the business perspectives that have been supported by conversational agents.
In the analysis of the studies to answer the general questions, we noticed the exploration of nine different business domains during the last decade. However, the majority of them focus on commerce, more specifically, e-commerce. This result may be due to the
Opportunities and future directions
Transparent and human-like conversational agents are determinant factors of the users’ trust during the dialogue [80]. One promising step towards these features should combine the three methods of the focused questions from this review, i.e., self-learning, personalization, and generative-based techniques. The self-learning allows to retrieve useful knowledge and proper features of users’ situation during the interaction. This knowledge should be preserved and maintained as historical
Conclusion
This article presented a systematic literature review about the use of conversational agents in business domains. The review answered three general, three focused, and two statistical research question based on a selected literature corpus. The corpus was created through the search of articles in research databases, with inclusion and exclusion criteria followed by filtering steps. Each article of the corpus was used to answer the proposed research questions, presenting the role of artificial
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
The authors would like to thank Dell Inc. for funding this work. We are also grateful to University of Vale do Rio dos Sinos (UNISINOS) and Applied Computing Graduate Program (PPGCA) for embracing this research.
References (83)
- et al.
Overview of the sixth dialog system technology challenge: Dstc6
Comput. Speech Lang.
(2019) - et al.
Recent advances on human-computer dialogue
CAAI Trans. Intell. Technol.
(2016) - et al.
Guidelines for conducting systematic mapping studies in software engineering: An update
Inf. Softw. Technol.
(2015) - et al.
Artificial conversations for customer service chatter bots: architecture, algorithms, and evaluation metrics
Expert Syst. Appl.
(2015) - et al.
Gamification and serious games in depression care: A systematic mapping study
Telemat. Inform.
(2018) - (2019)
How chatbots will transform the retail industry
(2018)- et al.
Automated assistance in e-commerce: An approach based on category-sensitive retrieval
- et al.
Predictive analytics in data science for business intelligence solutions
- et al.
A banking chatbot security control procedure for protecting user data security and privacy
Chatbot e-service and customer satisfaction regarding luxury brands
J. Bus. Res.
A survey on dialogue systems: Recent advances and new frontiers
SIGKDD Explor. Newsl.
AI-Based digital assistants
Bus. Inf. Syst. Eng.
A survey on conversational agents/chatbots classification and design techniques
Modern chatbot systems: A technical review
A survey on chatbot implementation in customer service industry through deep neural networks
Review of interactive communication systems for business-to-business (B2B) services
Electron. Imaging
How may i help you? – state of the art and open research questions for chatbots at the digital workplace
Anthropomorphic information systems
Bus. Inf. Syst. Eng.
Systematic mapping studies in software engineering
How to read a paper
ACM SIGCOMM Comput. Commun. Rev.
COBOTS - A cognitive multi-bot conversational framework for technical support
KNADIA: Enterprise knowledge assisted dialogue systems using deep learning
Transfer learning for context-aware question matching in information-seeking conversations in e-commerce
A new chatbot for customer service on social media
Conversational recommender system
Superagent: A customer service chatbot for e-commerce websites
Understanding chatbot-mediated task management
SARA: Singapore’s automated responsive assistant, a multimodal dialogue system for touristic information
Review-driven answer generation for product-related questions in e-commerce
Help yourself: A virtual self-assist system
Isa: Intuit smart agent, a neural-based agent-assist chatbot
Building task-oriented dialogue systems for online shopping
Leveraging conversational systems to assists new hires during onboarding
Lingke: a fine-grained multi-turn chatbot for customer service
Collection of resources and evaluation of customer support chatbot, vol. 307
SOGO: A social intelligent negotiation dialogue system
A multiagent-based technique for dialog management in conversational interfaces
An efficient search for context-based chatbots
A next generation chatbot-framework for the public administration
A hybrid approach to question-answering for a banking chatbot on turkish: Extending keywords with embedding vectors
Cited by (84)
Conversation-based hybrid UI for the repertory grid technique: A lab experiment into automation of qualitative surveys
2024, International Journal of Human Computer StudiesCan AI chatbots help retain customers? An integrative perspective using affordance theory and service-domain logic
2023, Technological Forecasting and Social ChangeAlert notifications for governmental disaster response via instant messaging applications
2023, International Journal of Disaster Risk ReductionService chatbot: Co-citation and big data analysis toward a review and research agenda
2023, Technological Forecasting and Social ChangeArtificial intelligence empowered conversational agents: A systematic literature review and research agenda
2023, Journal of Business ResearchMachine learning-based automation of accounting services: An exploratory case study
2023, International Journal of Accounting Information Systems