Chatbots: Overview and Enterprise Applications of AI-based Chatbots
Introduction
A chatbot is a computer program or service that simulates human conversation or, colloquially, human “chats”.1
While chatbots in the form of intelligent personal assistants like Siri, Alexa, Windows Copilot, and Google Assistant are popular among the public, an even larger enterprise market has emerged, facilitating what’s known as “conversational commerce,” and furnishing enterprise employees with new tools for data assimilation.
Digital commerce has entered its third generation:
- The first generation, electronic commerce, was marked by the development of websites – virtual storefronts in which products and services previously sold through hardcopy catalogs and “brick and mortar” stores could be acquired by anyone with access to a PC, a Web browser, and an Internet connection.
- The second generation, mobile commerce, was spurred by the development of smartphones and tablets and featured “mobile apps” which, in addition to websites, could be used to conduct commercial transactions.
- The third generation, conversational commerce, enabled by messaging platforms like Facebook and Twitter, permits users to communicate with businesses and other enterprise entities using natural language, either spoken via an audio interface or typed via text or other messaging app. The principal instrumentation of conversational commerce is the chatbot.
Chatbots have assumed greater relevance recently with the discovery that mobile users increasingly prefer to operate within messaging apps rather than social media sites. Likewise, they prefer to stay in messaging mode rather than exiting the environment to invoke single-purpose mobile apps.
The Chatbot Market
Grand View Research predicts that the global chatbot market, valued at $6.3117 billion in 2023, will reach $27.2972 billion in 2030, representing a respectable compound annual growth rate (CAGR) of 23.3 percent during the forecast period.
GVR attributes the projected rise to
- The need to provide 24×7 customers service.
- The need to reduce customer service costs.
- The demand among customers for self-service operations.2
Chatbot Basics
The Chatbot Concept
A chatbot is designed to answers questions or perform actions based on requests received via an audio, text, or other messaging interface, communicating with users in the same way that people communicate with each other. For example, instead of browsing an enterprise website in search of a particular product, a user could simply engage the enterprise’s chatbot, which would ask the user what she is looking for and, based on the user’s response, recommend one or more products that satisfy her requirements. The interaction is similar to what a user would experience if she visited the enterprise’s retail store and engaged a store clerk.3
Chatbots have been around for a long time. According to analyst Michael Yuan, “The very first chatbot, ELIZA, was developed [more than] 50 years ago at the Massachusetts Institute of Technology. It simulated a Rogerian psychotherapist, someone who just repeats the human user’s words back to the human.”4
External and Internal
There are two basic types of chatbots: external (customer-servicing) and internal (employee-servicing). As described by analyst Bharathi Ramadass:
- “External chatbots are domain-specific: They are designed to answer questions targeted to a particular topic such as customer service or healthcare.
- “Internal chatbots are the opposite. Think of them as a digital personal assistant that needs to process a wide range of topics typical of any employee experience.”5
Rule-Based and AI-Based
There are two basic types of chatbot operations: rule-based and AI-based.
- A rule-based chatbot is programmed to recognize and respond to specific keywords based on a set of prescribed rules. It has limited utility in situations where a user is imprecise in terminology or intent.
- An AI-based chatbot utilizes machine learning to infer what a user means. “It understands language, not just commands.”6
While rule-based chabots are less flexible than AI-based chatbots, they are also more predictable. AI-based chatbots, however, can improve themselves by detecting user behavior patterns, thereby increasing their utility and their popularity.7
Best AI-Based Chatbots
In a crowded market, ZDNET has identified its choices for the best AI-based chatbots:
Microsoft Copilot – Best AI chatbot overall. “Copilot features OpenAI’s most advanced LLM, GPT-4; has access to the internet; works like a search engine with information on current events; [and is] free.”
ChatGPT – Best original AI chatbot. “ChatGPT features OpenAI’s GPT-3.5 or GPT-4 (if subscribed); can generate text, solve math problems, and code; [has] impressive conversation capabilities; [and is] free to the public right now.”
OpenAI Website with Introduction to ChatGPT on Computer Monitor
Source: Pexels
Other, more specialized, selections include:
Anthrophic’s Claude – Best AI chatbot for summarizing documents.
Perplexity.ai – The best AI chatbot for prompt ideation.
Jasper – Best AI chatbot for businesses and marketers.
YouChat – Best AI chatbot that functions as a search engine.
Chatsonic by Writesonic – Best AI chatbot for article writers.
Gemini (formerly Google Bard) – Best AI chatbot if you’re a loyal Google user.
Socratic by Google – Best AI chatbot for kids and students.
HuggingChat – Best open-source chatbot.8
Enterprise Applications
The Employee’s Assistant
Enterprises have used chatbots for a number of years, usually to:
- Assist – or replace – customer service representatives.
- Assist customers visiting enterprise websites.
In addition to aiding individual employees, chatbots will be increasingly deployed to smooth and accelerate enterprise workflows, such as:
- Employee “onboarding,” helping familiarize new staff members with enterprise policies and protocols; and
- Basic IT support, permitting senior technicians to focus on non-trivial problems.9
The Salesperson’s Assistant
By leveraging their frequent interactions with customers and potential customers, enterprise chatbots can serve as critical repositories of marketing intelligence, including customer likes, dislikes, product preferences, and other data that can translate into all-important sales leads.
As analyst Abhiraj Dayal points out, beyond their routine data-gathering, “[Chatbots] can be programmed to ask specific pre-qualification questions and direct leads to the right team based on the responses for further nurturing. This would help in automating the sales funnel and allow sales reps to focus on more time-consuming tasks, such as closing deals!”10
Personalizing the Experience
Various studies have shown that “personalization” leads to an enhanced customer experience, which leads, of course, to greater customer satisfaction and more repeat business, thereby increasing a customer’s “lifetime value” – an especially relevant statistic given the generally high cost of customer acquisition. In other words, keeping current customers happy is normally more lucrative than gaining new customers. That’s where chatbots offer unique value.
As analyst Dayal explains,”[personalized] bots are able to remember and utilize information from prior conversations as this information is stored in a knowledge base. Truly personalized bots enable businesses to have a 1-to-1 conversation with each user, and they go beyond just focusing on demographics and product interests. By creating chatbots with personality and characteristics, organizations can enhance the customer journey and even increase conversion rates. Strong customer experiences are a necessity. Chatbot personalization is also a major advantage for diverse businesses that function in multiple verticals, as one chatbot can then have various personas for different audience segments.”11
Generative Artificial Intelligence
In an article entitled “The Chatbot Is Dead – Long Live the Chatbot,” Stephen Edison, Mike Evans, and Stephen Robnett of the Boston Consulting Group are bullish about the prospects for generative AI-based chatbots.
They observe, for example, that “Most consumers dislike [conventional] chatbots – less than a third of online customers use them – and for good reason. Chatbots’ mechanical language lacks a personal touch, they rarely understand the consumer’s specific needs, and offer generic and vague information. There’s not much real conversation taking place in chatbots today. For actual advice, consumers turn to offline, human interactions, and physical channels, creating cost for the business and extra friction for consumers.
Generative AI, however, “can transform online chat sessions into something more real,” helping realize the full potential of conversational commerce, and enabling retailers to “reduce the necessity for support by people.” GenAI users are similarly enthused. “According to a recent BCG survey of 500 US consumers, 66 percent expressed strong interest in trying GenAI-powered conversational commerce.”12
While the GenAI movement is accelerating, several major well-known problems remain:
- Accuracy, as GenAI can suffer “hallucinations”.
- Security, as GenAI systems are subject to cyber attacks.
- Privacy, as large language models (LLMs) on which GenAI trains can contain personally identifiable information (PII).
- Difficulty explaining results, as GenAI processes are frequently opaque.
- Copyright infringement, as LLMs often ingest information indiscriminately, and with little or no regard for intellectual property rights.
Usage Concerns
Inappropriate Language
In a recent scholarly paper entitled, “On the Dangers of Stochastic Parrots: Can Language Models Be Too Big,”13 the authors argue that “[the] tendency of training data ingested from the Internet to encode hegemonic worldviews, the tendency of LMs [language models] to amplify biases and other issues in the training data, and the tendency of researchers and other people to mistake LM-driven performance gains for actual natural language understanding – present real-world risks of harm, as these technologies are deployed.”
In an article called, “The Chatbot Problem,” analyst Stephen Marche contends that chatbots trained on large Internet data sets, can unknowingly adopt – and, more distressingly, express – the same biases and prejudices as many Internet users, leading to “unintended forms of racialized and gendered othering” speech.14
The lesson is: Watch what your chatbot is saying, and the language it is using.
Cyber Attacks
All forms of information technology – including chatbots – are vulnerable to cyber attacks and security exposures. To help mitigate the risk, analyst Mike Baker suggests that:
- Chatbot communication should be encrypted, and chatbots should be deployed only on encrypted channels.”
- Chatbot data should be properly protected. “Where this information is stored, how long it’s stored, how it’s used, and who has access to it must be addressed.”
- Enterprises should be on the lookout for criminal chatbots. “As chatbots become better at imitating humans, the technology will be used by hackers in phishing schemes and other social engineering hacks.”15
Personal Privacy
Audio chatbots, most commonly intelligent personal assistants (IPAs), are always listening. As analyst Deborah Matthews Phillips points out, this always-on aspect is potentially problematic, as sensitive or confidential information may be inadvertently – or intentionally – recorded by the chatbot provider.16
Customer Compliance
Research reveals what many chatbot analysts had anticipated and what many chatbot providers had feared. As Martin Adam, Michael Wessel, and Alexander Benlian report, “Though cost- and time-saving opportunities triggered a widespread implementation of AI-based chatbots, they still frequently fail to meet customer expectations, potentially resulting in users being less inclined to comply with requests made by the chatbot.”17
While AI-based chatbots are undoubtedly smart, they may not be smart-enough to conduct the highly random conversations that occur between an enterprise and its customers, particularly in a contact center context where customers may be resentful that they are not talking to a human agent.
Recommendations
The Chatbot Choice
Under the category of chatbot best practices, analyst Maddie Hoffman implores chatbot operators to:
Be Honest – “Tell users they’re interacting with a bot: Informing consumers that they’re conversing with a chatbot provides transparency, sets expectations, and promotes acceptance and trust.”
Be Accommodating – “Make it easy to reach a human agent: Sometimes, a customer may prefer to speak to a human agent, so you’ll want to give them the option to do so. It’s also important to ensure your bot can pass on the customer context and conversation history so agents have the necessary details.”18
Calculate Chatbot ROI
As with any IT product or service, enterprise officials should calculate the return on their chatbot investment. As detailed by analyst Abhiraj Dayal, there are three primary considerations in computing ROI:
- Query Resolution Time – The difference between the time taken by a live agent to resolve a particular problem or issue and the time consumed by a chatbot.
- Chatbot/Agent Work Dynamics – “Agents can only attend to a single user at a time. On the other hand, chatbots can handle multiple queries from various users at the same time and are available 24/7. The simple way to calculate this ROI would be to measure how much rudimentary work of one live agent could be handled by a chatbot.”
- Customer Satisfaction Level – Surveying customers relative to their chatbot versus agent experiences should produce an objective, i.e., numerical, ROI value.19
Assess Workforce Impact
As automated productivity tools, the wide scale use of chatbots will have a significant impact on employees in terms of:
- Worker training;
- Business process reengineering; and
- Even jobs, especially in HR, customer contact centers, and other departments where the impact of chatbots may be profound.
Enterprise officials should perform a Workforce Impact Assessment prior to any wholesale introduction of chatbots. To that end, the assessment process might be informed by conducting minor pilot programs, each testing the viability of a particular chatbot (or class of chatbots) within the enterprise ecosystem.
Web Links
References
1 Webopedia.
2 “Chatbot Market Size, Share & Trends, Analysis Report By Application (Customer Services, Branding & Advertising), By Type, By Vertical, By Region (North America, Europe, Asia Pacific, South America), And Segment Forecasts, 2023 – 2030.” Grand View Research, Inc. 2024.
3 Matt Schlicht. “The Complete Beginner’s Guide to Chatbots.” Chatbots Magazine. April 20, 2016.
4 Michael Yuan. “A Developer’s Guide to Chatbots.” IBM Corporation. August 10, 2016.
5 Bharathi Ramadass. “The Truth About Chatbots.” Forbes.com. January 21, 2022.
6 Matt Schlicht. “The Complete Beginner’s Guide to Chatbots.” Chatbots Magazine. April 20, 2016.
7 Jenna Alburger. “Rule-Based Chatbots vs. AI Chatbots: Key Differences.” Hubtype. July 25, 2023.
8 Sabrina Ortiz. “The Best AI Chatbots: ChatGPT Isn’t the Only One Worth Trying.” ZDNET, A Red Ventures company. February 16, 2024.
9 Team Linchpin. “25 Chatbot Stats and Trends Shaping Businesses in 2021.” Linchpin SEO, LLC. March 3, 2021.
10 Abhiraj Dayal. “How Chatbots Help Businesses to Improve Customer Experience in 2021.” Emplifi Inc. June 1, 2021.
11 Ibid.
12 Stephen Edison, Mike Evans, and Stephen Robnett. “The Chatbot Is Dead—Long Live the Chatbot.” Boston Consulting Group. December 15, 2023.
13 Emily M. Bender, Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell. “On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?” FAccT ’21: Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency. March 2021:610-623.
14 Stephen Marche. “The Chatbot Problem.” The New Yorker | Conde Nast. July 23, 2021.
15 Mike Baker. “What’s the Risk? Three Things to Know about Chatbots and Cybersecurity.” UBM. September 19, 2016.
16 Deborah Matthews Phillips. “Are Intelligent Assistants Smart Enough Yet?” Jack Henry & Associates Inc. August 17, 2016.
17 Martin Adam, Michael Wessel, and Alexander Benlian. “AI-Based Chatbots in Customer Service and Their Effects on User Compliance.” Springer. March 17, 2020.
18 Maddie Hoffman. “What Is a Chatbot?” Zendesk. February 21, 2024.
19 Abhiraj Dayal. “How Chatbots Help Businesses to Improve Customer Experience in 2021.” Emplifi Inc. June 1, 2021.