People’s Democratic Republic of Algeria
Ministry of Higher Education and Scientific Research
Saad Dahleb University
Blida - 1
Project Report
Field of Study: Computer Science
Specialization: Intelligent Systems Engineering (AI)
Theme
State Of The Art
Presented by
Benamara Ichrak
Messar Aya
Presented on: 22/02/2025
Academic Year: 2024/2025
State Of The Art
Contents
1 State of the Art 2
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 A Brief Historical Account for Chatbot . . . . . . . . . . . . . 2
1.4 Types of Chatbots . . . . . . . . . . . . . . . . . . . . . . . . 3
1.4.1 Based on Interaction Style . . . . . . . . . . . . . . . . 3
1.4.2 Menu/Button-based Chatbots . . . . . . . . . . . . . . 4
1.4.3 Keyword Recognition-Based Bots . . . . . . . . . . . . 4
1.4.4 Contextual Chatbots . . . . . . . . . . . . . . . . . . . 5
1.4.5 Hybrid Model Chatbots . . . . . . . . . . . . . . . . . 6
1.4.6 Comparison . . . . . . . . . . . . . . . . . . . . . . . . 6
1.5 Based on Response Generation Techniques . . . . . . . . . . . 7
1.5.1 Rule-Based Models . . . . . . . . . . . . . . . . . . . . 8
1.5.2 Retrieval-Based Models . . . . . . . . . . . . . . . . . . 8
1.5.3 Generative Models . . . . . . . . . . . . . . . . . . . . 9
1.6 Chatbot Architecture . . . . . . . . . . . . . . . . . . . . . . . 10
1.6.1 User Interface Component . . . . . . . . . . . . . . . . 11
1.6.2 User Message Analysis Component . . . . . . . . . . . 11
1.6.3 Response Generation Component . . . . . . . . . . . . 11
1.6.4 Natural Language Understanding . . . . . . . . . . . . 11
1.6.5 Context Tracking . . . . . . . . . . . . . . . . . . . . . 13
1.6.6 Dialog Manager . . . . . . . . . . . . . . . . . . . . . . 13
1.6.7 Backend . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.7 Internal Chatbot in Enterprise . . . . . . . . . . . . . . . . . . 15
1.7.1 Human Resources (HR) . . . . . . . . . . . . . . . . . 15
1.7.2 IT Support . . . . . . . . . . . . . . . . . . . . . . . . 16
1.8 How is an Enterprise Chatbot Different from a Normal Chatbot? 16
1.9 Challenges in Building Enterprise Chatbots . . . . . . . . . . 17
1.10 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1
1 State of the Art
1.1 Introduction
The rapid evolution of Artificial Intelligence (AI) and related technologies
has led to the development of innovative solutions that transform organiza-
tional processes and user experiences. Chatbots have emerged as powerful
tools for automating interactions and streamlining workflows. They can de-
liver instant and accurate responses to user queries using AI machine learn-
ing and Cloud-based services. This section explores the state of the art in
chatbot technology and focuses on using AI in organizational systems. A
review on technologies, methods, and challenges in the development of such
enterprise-focused chatbots.
1.2 Definition
A chatbot is a software application designed to simulate intelligent conver-
sation with users through either text or voice. It utilizes Natural Language
Processing (NLP) to understand and interpret human language inputs. By
processing the information provided to it, the chatbot can offer relevant re-
sponses or perform tasks such as performing calculations, setting reminders,
or activating alarms. [1] In the context of organizational systems chatbots
can be integrated with internal data sources to provide real-time access to
information such as employee questions, HR policies, marketing strategies
and business data. This integration enables chatbots to improve operational
efficiency by automating routine interactions and improving user experience
and enabling employees to focus on more strategic tasks.
1.3 A Brief Historical Account for Chatbot
The concept of chatbots dates back to the mid-20th century with early
attempts to simulate human-like conversations using machines. The evolu-
tion of chatbots can be summed in (figure1)
2
Figure 1: A brief history of chatbots.
In 1950, Alan Turing proposed the Turing Test as a measure of intelli-
gence for machines. The test involves a human judge conversing with both
a human and a machine via text, without knowing which is which. If the
judge cannot distinguish between the two, then the machine passes the test
and is considered intelligent ELIZA, the very first chatbot introduce prior to
the development of the first personalized computer. In 1966, Joseph Weizen-
baum developed ELIZA at the MIT Artificial Intelligence Laboratory (figure
1).According to a defined set of rules, ELIZA processes the keywords re-
ceived as input and then triggered the output responses. Several chatbots
still use this methodology of generating output. After ELIZA, PARRY came
relatively soon afterward called “ELIZA with attitude.” Standford Univer-
sity therapist Kenneth Colby process Parry, which stimulates an individual
with distrustful paranoid or paranoid schizophrenia. In 1995, ALICE or Al-
icebot, the inspiration of ELIZA, evolved by Richard Wallace. Although it
neglected to breeze through the Turing assessment, ALICE remained one of
the most rooted of its kind and honored with the Loebner Prize, an annual
AI competition, on several occasions.[2]
1.4 Types of Chatbots
1.4.1 Based on Interaction Style
Before the rise of the modern technology manual labor was the foundation
of operations across various industries. With technological advancements
chatbots have emerged as a valuable tool especially in areas like customer
service. However, chatbots are not a “one size fits all” solution. They dif-
fer significantly based on their user interface design, the algorithms driving
3
them and the technologies they use. They can generally be grouped into
three main categories (Figure2)
Figure 2: A brief history of chatbots. The first chatbot, ELIZA, was created
in the 1960s.
1.4.2 Menu/Button-based Chatbots
Menu/Button-based chatbots are a basic type of chatbots that are avail-
able today. Instead of the user having the ability to give free text input, the
user is limited to a number of buttons. They initiate the conversation with
the help of menus and dropdown menus, and they receive the response back.
These types of bots work on a mechanism of decision trees, and they direct
the users to select those options which will lead to the answer. Although
easy, these bots are slower in performance and may sometimes lack accuracy
or reliability in responses.[3] [4]
1.4.3 Keyword Recognition-Based Bots
Keyword recognition chatbots are an extension of menu/button chatbots,
the user gives a free text input which gets picked through a keyword. The
chatbot then analyses the text on specific keywords and based on those key-
words, gives an appropriate response. But they can face some troubles when
the keywords can overlap and create confusion. Example: A user asks this
query: How to enable auto-login feature in my mobile device? ”, the bot
may focus on keywords like ”auto” and ”Login” to deduce the best response.
4
Figure 3: chatbot type.
1.4.4 Contextual Chatbots
Contextual chatbots are text-based technologies used in the development
and testing of communication technologies. This also means that contex-
tual chatbots will use machine learning and AI-embedded technology, such
as voice recognition or sentiment analysis, to accomplish tasks. Instead of
just sending pre-programmed responses, it learns from the conversation his-
tory, understanding user intent and constructing intelligent, perceptive mes-
sages. For example, a contextual chatbot better workflows for Paid Time
Offs (PTOs)/Leave management operating a food delivery application may
use its previous ordering history, save payment and delivery addresses, and
suggest choices that best fit the user’s taste.
5
1.4.5 Hybrid Model Chatbots
The features of conversational agents driven by AI and menu-driven bots
are combined in hybrid chatbots. These bots are becoming more and more
popular since they provide users with convenience and flexibility. Some users
might not want to compose lengthy answers in real-world situations, therefore
hybrid chatbots offer an alternate method by posing a sequence of questions
with pre-selected answers. The experience is made more user friendly by
allowing users to participate in the manner of their choice.
Hybrid chatbots are used by several modern e-commerce platforms. To
improve interactivity and consumer happiness, these bots, for instance, might
engage customers in contextual conversations while presenting product cate-
gories and recommendations in a menu-driven fashion. A major development
in chatbot technology, hybrid chatbots provide a smooth fusion of intelligent
conversation and guided navigation.[5]
1.4.6 Comparison
A comparison between the types of chatbots is provided, highlighting their
advantages and limitations.
6
Chatbot AI/Rule Time for re- Advantages Limitations
Based? sponse giving
and training
Menu based Rule based Less training Quick response Limited usage.
less response Less ambiguity Only used for
time as paths are pre-
Yes/No type of
defined questions
Keyword Rule based More Training Easy for pre- More training.
Recognition More Response trained dataLimited set of
Time Fast response queries can be
answered
Contextual AI Based Requires signif- Learns from High develop-
chatbots icant training past interactions ment cost and
time to offer person- complexity
adaptive re- alized, requires large
sponse time intelligent re- datasets for
sponses training
Hybrid Chat- AI & Rule More Training Interactive Much time for
bots Based More Response Gives personal- training.
Time ized outputs Complex archi-
tecture
Table 1: Comparison of Different Chatbot Types
1.5 Based on Response Generation Techniques
There are two primary ways to produce responses: rule-based systems,
which use pre established patterns, and machine learning-based techniques,
which use data to build responses. More dynamic and context-aware replies
are provided by methods like Retrieval-Augmented Generation (RAG), which
have surfaced in recent developments. These techniques will be examined in
this section along with how they can enhance chatbot interactions.
7
Figure 4: chatbot type response generartion
1.5.1 Rule-Based Models
Rule-based chatbots follow a predetermined set of scripts and rules.
Their reasoning is based on a simple logical framework, and they usually use
a ”if-then” statement to determine their answers. [6]
key features:
• Use inflexible decision trees for operation.
• Derive answers based on specific keywords or inputs from users.
• Often simple to set up and cost-effective for performing simple tasks like
frequently asked questions—FAQs, or basic customer queries. Limitations:
Limited in flexibility, they can’t respond to questions that are outside their
set instructions. Interactions may even feel robotic, sometimes incapable of
providing a truly immersive experience for the user.[7]
1.5.2 Retrieval-Based Models
Retrieval-Augmented Generation (RAG) is a crucial part of our project’s
strategy for creating a useful and contextually aware chatbot. Since RAG is
essential to retrieving and producing responses from internal data, we will
give a technical description of it to highlight its significance and demonstrate
how it improves the operation of the chatbot. Retrieval-based chatbots make
sure to improve user interaction by finding the closest response to the user’s
input from a pre-defined set of answers. Unlike rule-based chatbots, which
work only on predefined rules, retrieval-based models use NLP techniques to
analyze and understand user queries.
Key Features:
• Response Selection: In place of generating new ones, retrieval-based chat-
bots select the best matching response from a database of pre-defined an-
8
swers.
• Natural Language Processing: The chatbots are based on NLP algorithms
that analyze users’ input. They are capable of finding the most appropriate
answer by discovering keywords, phrases, and the overall context in which
they appear.
• The semantic similarity and machine learning classifiers methods are mainly
applied in order to improve the accuracy of such a matching process.
• Curated Response Library: The success of a retrieval-based chatbot largely
depends on the quality and breadth of its response library. The library is
curated by developers to include a large number of possible questions so that
the chatbot is well-equipped to handle a variety of situations.
Limitations:
• Response Limitations: Although these chatbots can handle voluminous
amounts of queries, they remain limited to a set amount of predefined re-
sponses. Whenever a user goes beyond the limit with their question, the
chatbot may respond in a manner that would not make sense.
• Redundancy: Since the retrieval-based chatbots draw from a finite database,
there is also a repetitive nature when many users pose the same or simi-
lar information requests, not to mention the database becoming redundant
through a lack of diversity.[8]
1.5.3 Generative Models
The most sophisticated type of chatbots are generative ones, which use
machine learning (ML) and artificial intelligence (AI) to produce responses
on the fly. More fluid and organic interactions are made possible by these
bots’ ability to produce language that is human like depending on the cir-
cumstances of the conversation.
Features:
• Use neural networks and other deep learning models to comprehend and
produce language.
• able to modify responses in light of previous interactions and preserve con-
text after several exchanges.
• may gain knowledge from exchanges and get better with time as they en-
counter more data.
Limitations • Quality assurance processes can be challenging because an-
swers may sometimes be irrelevant or inappropriate.
• Dependence on large amounts of training data; poor training may result in
biased or insufficient responses.
• Require significant computational resources both during the training stage
and for operational activities, thus making their implementation costly. [9][6]
9
1.6 Chatbot Architecture
The architecture of a chatbot forms the backbone of functionality, en-
abling it to process user queries, generate responses, and interact with other
systems seamlessly. This involves integrating various components: natural
language understanding, dialogue management, and backend connectivity;
ensuring the chatbot can operate effectively across diverse use cases. A well-
structured architecture is one of the most important components of building
a scalable, efficient, and user-friendly conversational agent.
Figure 5: chatbot architecture
10
1.6.1 User Interface Component
The interaction of the chatbot starts when it gets an input provided
by the user in text or speech through a platform for communication. That
may be one of the currently popular messaging applications like Facebook
Messenger, Slack, WhatsApp, WeChat, Viber, or Skype.
1.6.2 User Message Analysis Component
The User Interface Controller acts as a bridge, channeling the user’s
input to the User Message Analysis Component. This component determines
the user’s intent and extracts relevant entities using pattern-matching tech-
niques or machine learning algorithms. Inputs can either be retained in their
original form—preserving grammatical and syntactical structures—or pro-
cessed using Natural Language Processing (NLP) techniques. In interacting
with a chatbot, users usually convey their goals or tasks through input, which
is referred to as ”intents.” Understanding such intents and executing corre-
sponding actions are core responsibilities of the chatbot. Inputs often contain
additional information, usually referred to as ”entities,” which provide fur-
ther context to perform precise execution. To improve accuracy, cognitive
services that can be integrated into the User Message Analysis Component
include: Spell Checkers: Correct user input to improve intent detection,
ensuring the input data is ”cleaned”. Machine Translation: Provide multi-
lingual interactions by detecting the user’s language and translating it into
the chatbot’s preferred language for processing through the NLP engine.
1.6.3 Response Generation Component
The Response Generation Component is responsible for generating
responses to user inputs, using one or more of three primary models: rule-
based, retrieval-based, and generative models. as previously mentioned each
model offers distinct advantages and limitations depending on the chatbot’s
use case and data availability.
1.6.4 Natural Language Understanding
Natural language understanding or NLU is the first core component of
the responding conversational agents about the providing of semantic repre-
sentation of a user utterance, [6] such as, in form of logic, an in-class intent,
it gives meaning to an utterance, [7]. Parsing, which takes the string of
words and provides a linguistic structure for the utterance. Implementation-
dependent is how an NLU uses to parse the input and can use context-free
11
Figure 6:
grammars or pattern matching or data-driven, as we see in the figure 2. What
NLU outputs needs to be able to be handle by a dialogue manager [8].
Figure 7: chatbot type.
Topic Detection
from figure2. Topic Detection is One of the most important steps in develop-
ing the conversational agents NLU stage is to identify the Topic of the text
(Conversation) for natural free-form interactions. Accurate tracking of the
conversation topic can be a valuable signal for a system for dialog generation
[9]. Examples on techniques can be used for topic classification are DANs
and ADANs.
12
Intent Analysis
An intent is a group of utterances with similar meaning. See the following
example of these two sentences: “I want to make a reservation in an Italian
restaurant” and “I need a table in a pizzeria”. They both have the same
meaning. That means the intent analysis goal is to identify the similarities
between words in meaning. If you can’t understand a user, your bot will
be useless whatever the effort you consume to develop the other components
such as dialogue management. How we can Teaching semantics to a machine.
We can do this about word vectors algorithms such as Word2Vec or Glove.
[10]
Entity Linking
the bottom component in figure2 consists of a Disambiguation model and
Named Entity Recognition(NER) and a template selection model. NER [11]
links entity mentioned in a text to a dictionary or knowledge base, local or
remote such as the whole web, to make sense of an entity and know what it is.
This is a significant step for allowing the chatbot to understand conversation
topics and generate appropriate responses. As it links the words of the user’s
utterances with concepts and subtexts in the real word. Options are to use
StanfordNLP, TAGME [12] or web mining via a search engine API. TAGME
can take the input text and returns a list of entities with their concept titles
from wikipedia, which in turn can be converted to nodes in the Wikidata
knowledge graph.
1.6.5 Context Tracking
When a sentence from a user appears, the chatbot gets this user’s most recent
utterances from the chat history database. Core reference can be resolved
using the Stanford CoreNLP toolkit [13]. If a coreferent was found, replace
pronouns and mentions of entities in the new sentence by.
1.6.6 Dialog Manager
The second core component in any chatbot, and which we can make a dif-
ference with other chatbots, is the Dialogue Manager, consisting of many
parts that could be improved or have additional ones in the future if that
will serve the DM. DM takes in user inputs from the NLU, producing at a
concept level the system responses to the natural language generator. The
strategy to be chosen, therefore, will depend on the response which the DM
will choose. Strategies are in relation to maintaining conversational state and
13
the ability to model the structure of dialogue beyond that of a single utter-
ance [6]. Being: rule-based, knowledge-based, retrieval based and generative.
The rule-based strategies being; backstory, intent templates and entity-based
templates ordered by their priorities. Since rule-based strategies encode hu-
man knowledge into the form of templates, they can give the most accurate
response. If the input is recognized by one of the strategies in the system, it
will adopt a template reply. Otherwise, it can try to get an answer from a
knowledge-base Q/A engine [5]. If that fails, a general conversation output
is achieved by an ensemble of neural network models along with information
retrieval modules.
1.6.7 Backend
It forms the backbone of the process that allows the chatbot to fetch informa-
tion required to address the intent of the user. This includes calling external
APIs or querying databases to access relevant data. Once the information
is fetched, it is sent to the Dialogue Management Module for processing and
passed further on to the Response Generation Module.
In the case of rule-based chatbots, the backend usually consists of a
Knowledge Base (KB) that has a set of predefined responses mapped to
certain user inputs. The Knowledge Base should cover a wide range of prob-
able user queries for effectiveness and provide variant responses for the same
input to avoid repetitive responses (Khanna et al., 2015).
Moreover, RDBs can be applied by chatbots in the storing and recalling
of past conversations. This makes a conversation relevant and continuous
since the bot refers to a previous conversation. The history conversation
access leads to more coherent and accurate dialogues that improve the user
experience A and John, 2015. It forms the backbone of the process that
allows the chatbot to fetch information required to address the intent of the
user. This includes calling external APIs or querying databases to access
relevant data. Once the information is fetched, it is sent to the Dialogue
Management Module for processing and passed further on to the Response
Generation Module.
In the case of rule-based chatbots, the backend usually consists of a
Knowledge Base (KB) that has a set of predefined responses mapped to
certain user inputs. The Knowledge Base should cover a wide range of prob-
able user queries for effectiveness and provide variant responses for the same
input to avoid repetitive responses (Khanna et al., 2015).
Moreover, RDBs can be applied by chatbots in the storing and recalling
of past conversations. This makes a conversation relevant and continuous
since the bot refers to a previous conversation. The history conversation
14
access leads to more coherent and accurate dialogues that improve the user
experience A and John, 2015
1.7 Internal Chatbot in Enterprise
Internal chatbots are designed in a way to facilitate internal workflows,
automate repetitive tasks, and improve overall productivity by providing
employees with on-demand assistance. These chatbots are built to support
multiple departments such as HR, IT, Knowledge Management, and many
others while ensuring seamless communication throughout an organization.
Organizations can also employ chatbots for internal communication to
increase support process [Link] the world witnesses chatbots on
websites, providing fast andautomated answers for online queries, they do an
equally important job at the back by helping out workers internally.
Key Use Cases for Internal Chatbots
Internal chatbots can be deployed across various organizational functions to
simplify tasks, improve communication, and provide timely assistance. Some
of the most common use cases include:
1.7.1 Human Resources (HR)
The core objective of the HR function is to add value to its primary
stakeholders, which include employees, managers, customers, and investors.
Driven by a growing demand to make strategic, bottom-line, and quantifiable
contributions to organizational performance while reducing costs, many HR
departments are now adopting technologies such as chatbots to facilitate
administrative processes in order for HR personnel to become more strategic
business partners and drivers of organizational change.
For example, chatbots are reported to add value by assist the HR func-
tion in recruitment processes, onboarding, training, and employer branding.
answering frequently asked questions-both inside and outside-on HR routine
tasks and improving employee engagement [10].
Better workflows for Paid Time Offs (PTOs)/Leave manage-
ment
The internal HR chatbots empower the employees to engage with the al-
ready existing HR and talent management software of their companies in
an easier way. The same user could avoid having to log in to a time-and-
attendance system to request time off by simply sending a natural language
message via channels such as Slack or Teams-”I need to request time off” or
”PTO”-and then have an HR chatbot prompt that continues to ask follow-up
15
Figure 8: PTOs
questions to hone in on precisely what kind of time off is needed. It then
confirms the details and processes the request, marking it in the company’s
attendance system in real time.
Employee Self-Service
The chatbots will answer frequently asked HR queries on benefits, leave
policies, payroll, and company policies, thus providing instant answers to
employees.
Employee Onboarding
Employees can get themselves introduced to company policies, fill in all
onboarding forms, and receive necessary training documents using chatbots.
1.7.2 IT Support
Help Desk Support: Chatbots can be used as the first line of contact
for IT support, helping users with general issues, troubleshooting steps, and
knowledge-based questions.
Incident Management: Chatbots help in managing incidents by capturing
the details, root cause identification, and initiating automated resolutions.
Change Management: Chatbots simplify change management by au-
tomating routine requests, approvals, and notifications.
Service Requests: Chatbots streamline service requests by allowing users
to submit, track, and resolve them through self-service platforms.
1.8 How is an Enterprise Chatbot Different from a
Normal Chatbot?
While an enterprise chatbot is specifically designed for the needs of an
organization, a normal chatbot would be developed to interact with a user
16
in general. Here are some of the key differences between an enterprise and a
normal chatbot:
Purpose
Enterprise Chatbots: These are business process-centric, can handle sev-
eral queries at one time, provide responses automatically, understand the
intent of the user to provide fast and accurate answers. They handle internal
workflows related to onboarding, application for leave, and tracking expenses
in order to make the processes faster and save employees’ time.
Normal Chatbots: They have very simple purposes, like giving informa-
tion or general queries.
Customization
Enterprise Chatbots: This is because, for business enterprise purposes,
these can be made highly customizable, featuring branding, multilingual ca-
pabilities, tailored greetings, and personalized user interfaces.
Normal Chatbots: Normally come with limited customization and, there-
fore, are better suited to general applications.
Security and Compliance
Enterprise Chatbots: With advanced security features like user authen-
tication and access control, they ensure that sensitive information is kept
safe. What’s more, they are compatible with the requirements of GDPR or
HIPAA, which enables the legal and ethical handling of information.
Normal Chatbots: They may lack in-depth security features and the ca-
pability for compliance since their development is not based on sensitive
information.
Advanced Natural Language Processing
Enterprise Chatbots: Equipped with rich NLP, such chatbots understand
even complex industry-specific terminology and jargon. They can handle
multiple languages, accents, and dialects, too, offering a more subtle sense of
what the user is trying to provide as input.
Normal Chatbots: In most cases, their capacity for NLP is confined to
simple conversations and general language comprehension.[11].
1.9 Challenges in Building Enterprise Chatbots
Developing enterprise chatbots comes with numerous technical hurdles.
A major challenge is ensuring that the chatbot can accurately understand
and process user inputs, which requires the use of sophisticated AI technolo-
gies such as Natural Language Processing (NLP). Additionally, integrating
chatbots with a company’s existing IT systems often presents difficulties,
demanding extensive customization, rigorous testing, and sometimes the de-
velopment of bespoke solutions [11].
17
User Experience and Design: Another big challenge is to create an
effective and interactive user interface for chatbots. The interface should
be designed in a way to let users navigate easily, maintaining the flow of
interaction so that users do not quit in the middle of a conversation. Besides,
the personalization of interactions to increase user trust and engagement is
highly important, but how to do it in a consistent and meaningful manner
remains a challenging task.
Data Privacy and Security: Since chatbots are mostly dealing with
sensitive information, one should be very concerned with privacy and secu-
rity issues in this regard. Any compliance with legislation on data protection,
like GDPR, is paramount for any organization, including solid security pro-
tocols to safeguard information breaches. Where security is kept, especially,
integrating your chatbot with numerous third-party services increases com-
plexity, and with increased levels of complexity, more vulnerabilities around
the protection of your data become open.
Data Synchronization: It’s one of the main challenges in integrating
chatbots with existing business systems. For the effectiveness of a chatbot,
information should be accurate and updated in real time.
Following are some of the problems that businesses usually face while
syncing data for integrating a chatbot:
Token size limitations: Large datasets may exceed the maximum token
size limit of some chatbot platforms.
Outdated information: Delays in syncing may result in the chatbot
providing incorrect or outdated data to the users.
Inconsistent data across systems: When multiple systems are in-
volved, data discrepancies can occur, causing confusion
1.10 Conclusion
The evolution of chatbot technology marks great strides in the automa-
tion of organizations and their communications. From simple, rule-based sys-
tems, modern chatbots have grown into complex AI-powered solutions that
can handle multiturn conversations, especially in an enterprise environment.
The review has shown that while current implementations successfully auto-
mate business processes across HR, IT, and knowledge management, future
developments will be related to context awareness, security, and enterprise
system integration. Success in deploying chatbots will remain pegged to
matching advanced capabilities with practical needs and security require-
ments.
18
References
[1] Anirudh Khanna, Bishwajeet Pandey, Kushagra Vashishta, Kartik
Kalia, Bhale Pradeepkumar, and Teerath Das. A study of today’s ai
through chatbots and rediscovery of machine intelligence. International
Journal of u-and e-Service, Science and Technology, 8(7):277–284, 2015.
[2] Rohit Tamrakar and Niraj Wani. Design and development of chatbot:
A review. ResearchGate, Apr, 2021.
[3] Aishwarya Gupta, Divya Hathwar, and Anupama Vijayakumar. Intro-
duction to ai chatbots.
[4] 3 chatbot types: Which is best for your business? — chatimize.
[5] Supreetha H V. A survey on various types of chatbots. International
Research Journal of Engineering and Technology, 2022.
[6] 5+ types of chatbots: From rule-based to ai.
[7] Rule-based vs. ai chatbots: Key differences - born digital.
[8] Chiranjib Chakraborty, Soumen Pal, Manojit Bhattacharya, Snehasish
Dash, and Sang Soo Lee. Overview of chatbots with special emphasis
on artificial intelligence-enabled chatgpt in medical science. Frontiers in
Artificial Intelligence, 6:1237704, 2023.
[9] Srijan Chatterjee, Manojit Bhattacharya, Sang Soo Lee, and Chiranjib
Chakraborty. Can artificial intelligence-strengthened chatgpt or other
large language models transform nucleic acid research? Molecular Ther-
apy Nucleic Acids, 33:205–207, 9 2023.
[10] Tina Taule. How can a chatbot support the hr function? exploring the
operational interplay a qualitative interview study. 2021.
[11] Enterprise chatbots: What are they, how to build them & more [guide],
2024. Accessed: Dec. 29, 2024.
19