ILEIID 2025 | International Journal of Research and Innovation in Social Science (IJRISS)

ISSN: 2454-6186 | DOI: 10.47772/IJRISS

Special Issue | Volume IX Issue XXV October 2025

Page 392

www.rsisinternational.org

AskEPTBot: Your Virtual EPT Chatbot Companion!

*Hairul Azhar Mohamad;

Muhammad Haziq Abd Rashid;

Hadayat Rahmah Hasan;

Nurhafeza

Mohd Akhir;

Amir Lukman Abd Rahman;

Urai Salam;

Sharifah Syazwa Amierah Syed Khalid

1,2,3,4,5

Akademi Pengajian Bahasa, Universiti Teknologi MARA, Shah Alam, Selangor, Malaysia

Fakultas Keguruan dan Ilmu Pendidikan (FKIP), Universitas Tanjungpura, Kota Pontianak,

Kalimantan Barat, Indonesia

Student Administration, Taylor’s University, Subang Jaya, Selangor, Malaysia

*Corresponding Author

DOI: https://dx.doi.org/10.47772/IJRISS.2025.925ILEIID000069

Received: 23 September 2025; Accepted: 30 September 2025; Published: 06 November 2025

ABSTRACT

AskEPTBot is a virtual innovation Artificial Intelligence (AI) chatbot by utilising the strengths of the

Telegram application system for diploma-level students and lecturers to seek information pertaining to the

English Placement Test (EPT). This innovative product was designed to replace the inefficient traditional

mode of communication to these stakeholders through the individual messaging platforms such as WhatsApp

or Telegram channels. Due to the repeated occurrences of misinformation and repetitive explanations given to

the interested parties, a one-stop chatbot centre was created to address the stated problems, thus reducing the

information overload and massive workload faced by the lecturers or known as EPT Lecturers in Charge

(LICs) who are assigned to relay continuous information to the students. The main aim of the innovation was

to reduce these burdens of administrative work through the automation of the communication process

involving frequent enquiries from 21 UiTM branches nationwide. This product can improve its responses

through more interactions with the users because it was integrated with machine learning. The analyses of

engagement metrics revealed that most interactions were centred around event-based enquiries such as

payment methods with 1,654 interactions compared to non-event or content related enquiries with high levels

of accuracy (94.61%). Meanwhile, the analysis of the user satisfaction survey indicated that 92 per cent of

users were satisfied with the clarity of information and 82 per cent of them were satisfied with the efficiency of

the chatbot. This AskEPT bot has a huge potential in terms of commercialisation through paid mobile apps

advertisement apart from its capacity to be upgraded into a web-based system that will attract more

advertisements from internal and external customers who want to promote their products. In terms of its

novelty, this AskEPTbot stands out as a one-stop centre hub for information about EPT with its additional

features of auto-promoted design of further related questions and adaptive ability for accuracy refinement. It is

recommended that this application is scaled up further with real-time content and dashboard-like reports for

administrative purposes to ensure its sustainable and profitable presence and usage for many years to come.

Keywords: AskEPTBot; EPT; AI Chatbot; Machine Learning; Commercialisation

INTRODUCTION

LCC111 (i.e., the newly revised label of its old code, ELC121) is a compulsory course in University

Technology MARA (UiTM) for diploma-level students. They will be eligible for exemption after passing the

English Placement Test (EPT). The information about this test has been communicated to the LICs and

students who will sit for the test through the traditional method of communication and EPT was before the

advent of AI-based Chatbot. AskEPTBot was developed on Telegram to help consolidate information about

EPT in a centralised manner to support the administrative process of the test by giving auto-generated answers

ILEIID 2025 | International Journal of Research and Innovation in Social Science (IJRISS)

ISSN: 2454-6186 | DOI: 10.47772/IJRISS

Special Issue | Volume IX Issue XXV October 2025

Page 393

www.rsisinternational.org

to common EPT questions, hence streamlining the communication process. This critical shift to the virtual

information delivery

model centred on the students managed to resolve numerous challenges related to the EPT administration.

Initially, all information about EPT was fragmented in many ways and platforms on text messaging

applications between the test administrators and the students, thus causing a lack of streamlined details,

duplicate promotional documentations, missed notices, and misinformation. According to Romero-Saritama et

al., (2025), WhatsApp is one of the most popular online networking tools in the formal and informal settings,

but it is presented with information overload because of uncontrollable traffic. Baishya and Maheshwari

(2022) further adds that having

a centralised automated channel can reduce the challenges with information delivery pertaining to important

dates, process and procedures, and FAQs, especially during high-stake situations or events. Therefore, the

development of AskEPTBot addresses the needs and benefits of the situations.

Chatbots used by universities can effectively decrease the administrators’ and lecturers’ workload through the

chatbots’ improved responsiveness to students’ common questions. These efficiency in chatbots can be

attained with the system’s better curation and clever escalation pathways. As argued by Labadze et al. (2023),

the reports on the increased productivity in higher education were attributed to the efficient use of

communication on the condition that governance and accuracy of information are given due consideration, in

which AskEPTBot is striving to achieve by being a one-stop information centre to the students in the form of

continuous automated chatting experience anytime and anywhere. Zendesk (2025) clarified that when

repetitive questions are handled by a specific automatic system such as a machine-assisted channel, the staff

can refocus on more complex and unique cases. This met the aim of the development of AskEPTBot to

increase the quality of handling the pre, during, and post-test administration.

Objectives

This innovation sought to reduce the burden and workload of the lecturers in charge (LICs) as well as

administrators in UiTM in handling the enquiries from students nationwide pertaining to all types of common

topics. This involves thousands of enquiries from 21 UiTM branch campuses nationwide. As a result, this

helps alleviate their lecturers’ worries and concerns about incorrect and unupdated information they commonly

had to respond to through WhatsApp. By creating an automated machine-assisted question and answer (Q&A)

chatbot system, this allows their energy and time to be allocated to the EPT administration and handling

process as well as time-sensitive test matters.

METHODOLOGY & PRODUCT DESCRIPTIONS

The development of AskEPTBot underwent the intricate and detailed process of understanding the capabilities

of Telegram to be an automated chatbot to the students. By leveraging on the Telegram API always-on access,

AskEPTBot was developed to be the first Telegram-based one-stop information centre chatbot to the students

of UiTM involving test administration. Hence, it is named after the test itself - English Placement Test (EPT).

It was created to assist with enquiries about EPT which included the test participation pre-requisites, the

important dates and events the potential candidates need to know, the test registration and payment methods,

the trial test resources before the real test days, tutorials pertaining to payment getaway methods, proof of

payment and the procedures of sitting for the test conducted on the university’s online student portal platform

known as UFuture, and the available slots to take the tests involving 9 to 12 slots for 3 to 4 test days to be

chosen by the thousands of students and assigned by the LICs, the announcement of the test results, the

application of the first level English diploma code after passing the test and many more.

This chatbot provides simple and detailed information depending on the further prompts entered onto the

chatbot prompt box. Due to large amounts of information handled by the system, it had gone through several

trial test runs. Its conversational interface and automated responses allow for unlimited messaging experience

to the students. The chatbot system is designed to have clear and quick information retrieval with structured

ILEIID 2025 | International Journal of Research and Innovation in Social Science (IJRISS)

ISSN: 2454-6186 | DOI: 10.47772/IJRISS

Special Issue | Volume IX Issue XXV October 2025

Page 394

www.rsisinternational.org

menus for the users to choose from after every response. This ensures a continuous guidance and dependable

interaction during all situations especially when the students register for the test, make the fee payment, access

trial resources, and result information. (Telegram, n.d.; UiTM Library, 2024).

To ensure clarity and accuracy of the information, the system developer was given details regarding the

inquiry categories and types of inquiries.

Table 1: Enquiry Category and Type of Enquiry

Enquiry Category

Type of Enquiry

Significant Event-based Enquiries

on AskEPT Chatbot

Telegram Group Link Sharing

Payment Procedures

Payment Status Validation

Trial Test Paper Access

Significant Non-Event Enquiries

(General Inquiries) on AskEPT

Chatbot

EPT Test Format

Tutorials on EPT

Important Dates for EPT

Introduction to EPT

As shown in Table 1, there are two broad types of enquiry categories. The development of the content was

based on a knowledge-base-first approach by curating and selecting the most common information and

mapping it into two types of intents: 1) event-based and 2) non-event. When an enquiry is prompted, this

creates a potential flow of interactive chats within a window period of two minutes. If left idle with no follow-

up enquiries after the first response, the session will be automatically renewed into a new chat.

The chatbot system will detect if the first enquiries falls in either of the two broad categories to suggest

prompts with more specific possible enquiries to the users. Event-based enquiries include time-related tasks

such as registration period, payment methods, payment validation process, trial test resources access.

Meanwhile, the non-event category entails questions about common FAQs (test format, tutorials, dates, and

brochures and marketing materials about EPT, etc).

In addition, AskEPTBot comes with detailed reports that can be generated manually from the back-end

operation system. This includes engagement metrics such as number of users, number of sessions and

interactions, containment periods, and response time, apart from interactive optimisation information

(Intercom, 2023; Quickchat AI, 2025). These metrics were selected to measure the engagement details shown

by AskEPT users because the metrics offer a detailed and complete analysis of the users’ activity in terms of

entering to use the AskEPTBot, the frequency of interactive prompts, and the length of their interaction.

To check whether the use of AskEPTBot is helpful and effective to the users, a specific satisfaction survey was

used. The survey was an existing survey adopted from the past study which has been validated to have fulfilled

the adequate psychometric requirements. It is a questionnaire survey with Likert-scale items to measure user

perceptions of AskEPTBot Interactiveness, Content Precision, and Interaction Speed. The questionnaire was

pre-tested with a small group of students to ensure clarity and reliability of the items (Silva & Canedo, 2022).

Thus, AskEPTBot was equipped with this usefulness survey questionnaire, which is a simple and easy-to-rate

user experience survey embedded into the last end of the chat thread accessible through a different auto-

prompted platform throughout the accessible period of the AskEPTBot. The survey is important to see the

users’ perceptions of the AskEPTBot’s levels of information clarity, speed, interaction pace, appropriateness,

ILEIID 2025 | International Journal of Research and Innovation in Social Science (IJRISS)

ISSN: 2454-6186 | DOI: 10.47772/IJRISS

Special Issue | Volume IX Issue XXV October 2025

Page 395

www.rsisinternational.org

and the users’ expectations as well as their overall user experience and satisfaction levels towards the quality

of the chatbot usage. According to Kommunicate (2025), this type of survey is important to be integrated to

evaluate the product capabilities and changes in terms of the user’s engagement with any observed frictions.

FINDINGS & COMMERCIALISATION

Figure 1: Finding on User-Interaction Ratio for One Month

Figure 1 above shows the finding based on the engagement metrics analysis on the User-Interaction Ratio in

one month period. One user can generate one or more sessions in which one session contains one or more

interactions. An interval duration for a 2-minute inactivity automatically redirects the users to a fresh new chat

session, closing an earlier one as one completed session. The engagement data revealed that there are 5,067

users who interacted in multiple sessions within this duration. A total of 24,279 sessions with 27,721

interactions, thus generating a user-session-interaction ratio to a ratio of 1:4:5. This increasing pattern suggests

a returning usage pattern within a brief window period among users which were typical for situations related to

deadlines in which AskEPTBot was revisited to create chained interaction tasks compared to single interaction

queries. According to a report by Quickchat AI (2025), these behavioural patterns of sessions per user are

consistent with normal and healthy engagement levels involving multi-step flows of generated interactions

when human intervention is not included.

Figure 2: Finding on Overall Users for Event and Non-event Based

Figure 2 shows the finding based on the engagement metrics analysis on the overall users’ enquiries in terms

of event-based and non-event-based enquiries. It was discovered that event-related enquiries with 5,032 units

ILEIID 2025 | International Journal of Research and Innovation in Social Science (IJRISS)

ISSN: 2454-6186 | DOI: 10.47772/IJRISS

Special Issue | Volume IX Issue XXV October 2025

Page 396

www.rsisinternational.org

of occurrences surpassed the non-event enquiries with 3,705 units of occurrences within the observed period.

The difference between the two types of enquiries is 1,327 counts. If converted into percentages, event-based

enquiries made up 57.60 per cent compared to non-event ones making up the remainder of 42.40 per cent. This

finding suggests that time-bound situations such as registration, payment, and validation periods prompted

more enquiries from the users than general information-related enquiries such as information about EPT and

its requirements. According to a report by Intercom (2023), time-constraint information has the plausibility of

driving the triggers for more enquiries from users. According to Zendesk (2025) pertaining to the perspective

of system design, displaying the system dialogs to be centred towards important events increase the first-

attempt resolution, thus reducing follow-up interactions from users, making this pattern a most widely

accepted style of information design in AI-assisted environments.

Figure 3: Finding on Engagement Rate (Event-Based)

Figure 3 shows the finding from the engagement metric analysis on the types of specific enquiries related to

event-based category. It was revealed that event categories recorded high volumes for Payment Procedures

(n = 1,654), Trial Test Paper Access (n = 1,583), Payment Status Validation (n = 1,262), and Telegram Group

Link Sharing (n = 533). It can be suggested that the dominant record of enquiries on payment matters indicate

the significant impact of having an automated payment system in which the link is retrievable from the

AskEPTBot compared to enquiries pertaining to the various telegram links created to cluster potential

candidates into specific channels based on their branch campus administered by their campus LICs for more

targeted information sharing.

Figure 4: Finding on Engagement Rate (General Enquiries)

Figure 4 shows the findings from the engagement metrics analysis on engagement rate for non-event intents

(general enquiries). Enquiries focusing on EPT Test Format (n = 1,061) recorded the highest volumes

ILEIID 2025 | International Journal of Research and Innovation in Social Science (IJRISS)

ISSN: 2454-6186 | DOI: 10.47772/IJRISS

Special Issue | Volume IX Issue XXV October 2025

Page 397

www.rsisinternational.org

compared to enquiries focusing on Tutorials (n = 959), Important Dates (n = 912), and Introduction to EPT

(n = 773). This data indicates that information related to the test format is rather unfamiliar to the users, hence

generating the most volumes of enquiries for a proper orientation before sitting for the tests. This information

is considered timeless details the student users need to be equipped with to prepare themselves in advance,

hence having a chatbot that can answer standard enquiries on EPT reduces numerous burdens and brings about

a great workload relief to the EPT administrators and LICs. The scaffolding system remains a continuous and

important driver to the development of self-service chatbots in higher education if the details are consistently

updated and prices (Labadze et al., 2023; Gökçearslan et al., 2024).

Figure 5: Finding on Engagement Rate (Response Accuracies)

Figure 5 shows the finding from the engagement metric analysis on the response accuracies of the engagement.

Based on a one-week sampled data, it was reported that AskEPTBot managed to achieve 94.61 per cent

accurate responses in providing information to the users. This left the inaccurate responses with 5.39 per cent

only. This is due to the strong machine learning performance of AskEPTBot in improving its clarity,

corrections, citation practices based on the old data provided by the users, making it an important aspect of an

automated chatbot technology (Yau et al., 2024; Shiferaw et al., 2024). In addition, routine updates and audits

with retrieval capacities for high-stake situations assisted with the accuracy of the information supplied to the

users, thus preserving trust and confidence for follow-up reconnections and contacts during numerous

occasions.

Figure 6 & 7: Finding on Usefulness Perception towards “Easy to Understand Response” & “Find Chat

Response Slow”

ILEIID 2025 | International Journal of Research and Innovation in Social Science (IJRISS)

ISSN: 2454-6186 | DOI: 10.47772/IJRISS

Special Issue | Volume IX Issue XXV October 2025

Page 398

www.rsisinternational.org

Figure 6 and 7 show the findings based on a usefulness survey on two specific questions in the usefulness

survey: 1) “Easy-To-Understand” and 2) “Find Chat Response Slow”. It was revealed from the survey

responses that users indicated the ease to understand the content of AskEPTBot with high ratings (251/454

times or 55.3 per cent and 167/454 or 36.8 per cent) compared to the negative options (“Sometimes”, “No,

most of the time”, “No, every time”), making their positive ratings to score a total of 92.1 per cent. This

finding suggests that AskEPTBot answers obtained by the users are highly understandable to them.

Meanwhile, responses to the survey item on the slowness of the chat responses revealed that users expressed

their disagreement with this survey item. The users disagreed that AskEPTBot generated slow responses with

the every-time disagreements from 193/454 of them or 42.5 per cent and most-of-the-time disagreements from

143/454 of them or 31.5 per cent, making a total of 74 per cent disagreements. This finding suggests that

AskEPTBot are perceived to be efficient in providing quick answers to their enquiries. These perceptions are

supported by Intercom (2023) and Zendesk (2025) that clear and quick responses are two critical drivers of the

usefulness of an AI chatbot system in their conversational experience with the users. This pattern also implies

that easy-to-follow and efficient content integrated in the copywriting of chatbot answers help increase the

perceived utility of the chatbot product during their system-user conversations.

Figure 8 & 9: Findings on Usefulness Perception towards on “Get Appropriate Interaction Pace” & “Get

Appropriate Response”

Figure 8 and 9 show the findings from the usefulness survey on the perceptions of the users towards two more

items: 1) “interaction pace was appropriate” and “responses were appropriate”. It was found that the

student users perceived the appropriateness of the interaction pace with 257/458 or 56.1 percent of them

agreeing this to happen every time and 160/458 or 30.4 per cent agreeing this to happen most of the time. The

combined two top two ratings generated a total of 417/458 or 91.04 per cent indicating their favor towards this

perception. This finding suggests that AskEPTBot has the suitable pace of interaction and responses in line

with their prompting speed.

Furthermore, they also showed that they received appropriate responses with 277/458 or 61.0 per cent of them

agreeing this to happen every time they prompt and 138/454 or 30.4 per cent of them agreeing this to happen

most of the time. The combined percentage of the top two ratings is 415/458 or 90.6 per cent expressing their

overall favor towards this item. It can be suggested from the finding that AskEPTBot could provide relevant

information that can help them address their concerns.

These conclusions are aligned with Al-Shafei (29025) and Ladadze et al. (2023) that effective conversation

pace and response fit increases the satisfaction of the users to repeat their usage of AI Chatbot. Therefore,

ILEIID 2025 | International Journal of Research and Innovation in Social Science (IJRISS)

ISSN: 2454-6186 | DOI: 10.47772/IJRISS

Special Issue | Volume IX Issue XXV October 2025

Page 399

www.rsisinternational.org

designing response interaction speed and content suitability in AI Chatbot helps sustain the flow of interaction

through a multiple-step process in educating the users with information needed.

Figure 10 & 11: Findings on Usefulness Perception towards “Meet Expectation” and “Overall Efficiency

Rating”

Figure 10 and 11 show the findings from the usefulness survey on the perceptions towards two more items: 1)

“Meet Expectations” and 2) “Overall Efficiency”. It shows that 202/454 or 44.5 per cent of the respondents

agreed that the AskEPTBOt met their expectations every time and 189/458 or 41.6 per cent agreed most of the

time that this chatbot met their expectation. This finding suggests that AskEPTBot managed to generally meet

the expectations and needs of the student users in delivering their enquiries.

In addition, the perception survey also revealed that 373/454 or 82.2 per cent of the respondents were satisfied

with the overall efficiency of AskEPTBot every time they used the conversational application. Meanwhile,

80/458 or 17.6 per cent agreed most of the time that the application managed to address their questions. The

combined percentage of the two top ratings is 99.8 per cent, indicating a huge majority of the student

respondents were generally happy and satisfied with the usage of this chatbot application in providing

information they need. It can be suggested from both findings that the application did serve the purpose and

intention well to the users. This conclusion is consistent with Stockholm University (2024) and EdTech

Magazine (2025) on the importance of student chatbot to be designed for its purposes based on its needs in

higher education contexts so that its follow-up usage will be sustainable in the long run.

Based on the findings from the engagement metrics and the useful perceptions survey, the project assisted with

the next potential course of action regarding the commercialisation values of AskEPTBot. There are two

commercialisation values that can be potentially enjoyed from this AskEPTBot.

Commercialisation Value 1: Monetising from download (mobile application)

The first commercialisation value is monetisation from downloads on the upgraded mobile application. The

AskEPTBot, once upgraded into its own mobile application version, can include a freemium-with-subscription

model. In the application version. For student users who feel that they need to get more EPT-selective

resources and more specialised materials who can help them excel in the test, they will be able to do that at a

cheap price. As substantiated by AppsFlyer (2024), this method will help the faculty to generate income which

proves to be common in the industry of mobile applications. This method is acceptable if the educational

materials application monetizes moderately and reasonably that can be used to fund a continuous upgrading of

the application for a better and greater immersive experience with the chatbot and its application. Median

ILEIID 2025 | International Journal of Research and Innovation in Social Science (IJRISS)

ISSN: 2454-6186 | DOI: 10.47772/IJRISS

Special Issue | Volume IX Issue XXV October 2025

Page 400

www.rsisinternational.org

(2024) maintains that this strategy and form of monetisation is the resilient approach to the system

administrators which is substantially growing in non-gaming realms. For EPT situations, apart from allowing

free downloads for standard content in ensuring fairness to other users, paid downloads with a light fee could

also allow offline access and priority access during strict or high-traffic situations.

Commercialisation Value 2: Monetising from advertisements (web-based)

Another commercial value to be drawn from this AskEPTBot is the monetisation of the advertisements from

the advertisements on the free version of the applications and its website. The student users without

subscription can still use the application, but with some limitation in terms of the compulsory views of the

advertisements before they can get access to the materials and selected applicable materials.Two levels of

selective materials which are meant for add-view versions and full paid version of the upgraded plans. The

usage of in-app or web advertisements as a form of monetisation continues to be the way the advertisement

administrator e employs in their applications worldwide. According to Statistica (2025), it is projected that this

method of monetisation will reach nearly USD390 billion and Oberlo (2024) indicate that this will be a steady

growth. In the educational context, AskEPTBot may employ non-intrusive advertisement policy with clear

information indication and disclaimers about the usage of the application prior to the application installation.

SUMMARY & CONCLUSION

It can be concluded from the findings and commercialization discussion that AskEPTBot has high interaction

volumes among its student users in which on average one student had a minimum of 4 sessions and five

interactions. Most enquiries were centred on the questions related to event-based type compared to general

enquiries with almost 95 per cent accuracy of answers given to them. In terms of survey, student users

expressed their ease of understanding, speed of response, appropriateness of response, appropriateness of the

interaction pace and answers meeting their expectations. These high ratings towards these aspects are also

reflected in their overall rating of the usefulness of AskEPTBot to them. Sustained and moderate gains can be

obtained from the commercialisation of AskEPTBot through the paid downloads of the upgraded version to

application-based product as well as the advertisements posted on the application and its website, thus making

these primary propositions for the continuous and sustainable development of AskEPTBot.

NOVELTY & RECOMMENDATIONS

There are several novelties that can be identified with AskEPTBot and recommendations of its future version

and usage. These novel and recommended features will help propel AskPTBot to a different level to the future

student users.

Novelty 1: One-stop centre (knowledge base app-based tool)

The first novel feature is that AskEPTBot is positioned to be a one-stop centre for information and knowledge

to the EPT-interested students without causing them to be bombarded with unstructured, unfiltered, repetitive,

and fragmented information from various sources and persons in charge. AskEPTBot has successfully

consolidated their materials required weeks before and after the test administration. It is also accessible in a

conversational manner through a specific telegram channel handle that can be used not only by the students but

also the other administrative staff who act as the connecting points to the students. As a result, the staff can

focus their concerns and assistance on more critical cases or unique situations with better understanding of the

structure and the details of EPT. In addition, the integrated model has lowered the administrative cost and

effort previously endured through the traditional or conventional methods. According to Labadze et al. (2023)

and Gökçearslan et al. (2024), this method of communication and integration is a great approach to

centralisation leading to faster resolutions and reduction of the staff workload with improved maintenance of

its content revision and versioning.

Novelty 2: Auto-prompted questions and auto-responses

ILEIID 2025 | International Journal of Research and Innovation in Social Science (IJRISS)

ISSN: 2454-6186 | DOI: 10.47772/IJRISS

Special Issue | Volume IX Issue XXV October 2025

Page 401

www.rsisinternational.org

Furthermore, AskEPTBot is unique with its auto-prompted choices of the follow-up enquiries that the users

may be interested in. Auto-prompted choices reduce ambiguity by increasing the speed in capturing their intent

for first-attempt resolutions, while auto-responses reduce critical phrasing times such as during peak hours of

usage activity. Research shows that structured dialog and service scripts improve users’ satisfaction levels

especially when clear escalation rules and empathetic tone are provided (Al-Shafei, 2025; Intercom, 2023).

This pattern decreases cognitive thinking load and avoids unnecessary back-and-forth interactions.

Novelty 3: Machine learning for accuracy gains with engagement

Next is the novel feature of machine-learning technology. Speedy information retrieval models driven by

machine learning technology due to the growing usage of AskEPTBot helps improve response accuracy and

coverage with sustained essential quality control. According to Yau et al. (2024), continuously recent

evaluations of automated system information underscore the importance of accuracy monitoring, reducing

delusional information risks, and validation of domain sourcing before responses are given to the users, thus

sustaining their authenticity. Thus, AskEPT is well-equipped with such learning capabilities which are

consistently incremental apart from being paired with proper and strict governance of the chatbot system

administration.

RECOMMENDATIONS

It is recommended for future improvement of the usage of AskEPTBot that the commercialisation of this

chatbot include more precise forecasts or results from pilot trials to substantiate its extended practicality. This

can be done by adopting a pilot approach through freemium-first with test light subscriptions for specific times

(AppsFlyer, 2024; Statista, 2025). AskEPT is seasonally conducted before each new semester begins.

Therefore, such an implementation can be done in phases involving selective events, certain focused intents,

and moderate-stake information instead of abrupt overall implementation. By common practice, specialised

chatbot gives a greater priority to high-traffic intents (payment, trial access, validation) for answer

consolidation by including sources, citations, visuals, and screenshots (Yau et al., 2024; Stockholm University,

2024). Thus, this detailed information will assist in increasing clarity to the users. According to Intercom

(2023) and Quickchat AI (2025), the chatbot system is highly recommended to have quality dashboards and

reports with the purpose of tracking accuracy, containment, and errors, and fallbacks. AskEPTBot may have

such features through weekly or monthly reviews. before EPT windows.

Adding a deeper examination of possible challenges, which include scalability across different contexts and

concerns about data privacy, would provide essential balance. This can potentially be done by expanding

evergreen content such as updated format, more informative tutorials, and more specific dates with structured

playbooks or short slips is helpful for the improvement of the system (Labadze et al., 2023). In the case of

AskEPTBot which is seasonal in nature, this can be implemented for the new cohorts and done in batches. For

risky, unresolved or sensitive enquiries, Gökçearslan et al. (2024), the chatbot can integrate system-driven

escalation triggers to ensure the queries are not left unattended. This allows the AskEPTBot administrators to

be notified of such situations and take necessary actions outside of the chatbot system. In future, such triggers

can be considered in the next round of system updates to be included as part of the input for future automation

processes.

Finally, a competitive comparison with other educational chatbot systems would also help position

AskEPTBot within the broader academic discourse. Such an analysis could highlight unique strengths and

weaknesses, offering insights into areas where the system excels or requires improvement for a greater

academic impact. Additionally, this comparison would provide readers with a clearer understanding of how

AskEPTBot aligns with current trends and innovations in educational technology.

REFERENCES

1. Al-Shafei, M. (2025). Navigating human-chatbot interactions: Factors influencing user satisfaction and

engagement. International Journal of Human–Computer Interaction, 41(1), 411–428.

https://doi.org/10.1080/10447318.2023.2301252

ILEIID 2025 | International Journal of Research and Innovation in Social Science (IJRISS)

ISSN: 2454-6186 | DOI: 10.47772/IJRISS

Special Issue | Volume IX Issue XXV October 2025

Page 402

www.rsisinternational.org

2. AppsFlyer. (2024). The state of app monetization—2024 edition.

https://www.appsflyer.com/resources/reports/app-marketing-monetization/

3. Baishya, D., & Maheshwari, S. (2022). WhatsApp groups in academic context: Exploring the academic

uses of WhatsApp groups among students. Contemporary Educational Technology, 14(1), 1–12.

https://www.academia.edu/84799466/ WhatsAppGroupsinAcademicContext (preprint)

4. EdTech Magazine. (2025). How three universities developed their chatbots.

https://edtechmagazine.com/higher/article/2025/05/how-three-universities-developed-their-chatbots

5. Gökçearslan, Ş., Tosun, C., & Erdemir, Z. G. (2024). Benefits, challenges, and methods of AI chatbots in

education: A systematic review. International Journal of Technology in Education, 7(1), 19–39.

https://files.eric.ed.gov/fulltext/EJ1415037.pdf

6. Intercom. (2023). How are customer service metrics changing in the age of AI?

https://www.intercom.com/blog/customer-service-metrics-ai/

7. Kommunicate. (2025). Metrics and attributes used in Zendesk Chat.

https://www.kommunicate.io/blog/metrics-and-attributes-used-in-zendesk-chat/

8. Labadze, L., Grigolia, M., & Machaidze, L. (2023). Role of AI chatbots in education: Systematic literature

review. International Journal of Educational Technology in Higher Education, 20, 56.

https://educationaltechnologyjournal.springeropen.com/articles/10.1186/s41239-023-00426-1

9. Median. (2024). Mobile app monetization statistics 2024: Trends & earnings insights.

https://median.co/blog/mobile-app-monetization-statistics-2024-trends-earnings-insights

10. Oberlo. (2024). In-app advertising revenue (2017–2028). https://www.oberlo.com/statistics/in-app-

advertising-revenue

11. Quickchat AI. (2025). The complete guide to chatbot analytics: KPIs, dashboards, and deflection rate.

https://quickchat.ai/post/chatbot-analytics

12. Silva, G. R. S., & Canedo, E. D. (2022). Towards User-Centric Guidelines for Chatbot Conversational

Design. International Journal of Human–Computer Interaction, 40(2), 98–120.

https://doi.org/10.1080/10447318.2022.2118244

13. Statista (In-App Advertising—Worldwide). (2025). Market outlook.

https://www.statista.com/outlook/amo/advertising/in-app-advertising/worldwide

14. Stockholm University—Department of Education. (2024). New research highlights opportunities and

challenges of AI chatbots in higher education. https://www.su.se/department-of-education/news/new-

research-highlights-opportunities-and-challenges-of-ai-chatbots-in-higher-education-1.759929

15. Telegram. (n.d.). Telegram Bot API. https://core.telegram.org/bots/api

16. UiTM Library (PTAR). (2024). English Placement Test (EPT) event.

https://library.uitm.edu.my/en/component/jem/event/325-english-placement-test-ept

17. Yau, J. Y.-S., et al. (2024). Accuracy of LLM chatbot responses for emergency care questions. Journal of

Medical Internet Research, 26, e60291. https://www.jmir.org/2024/1/e60291

18. Zendesk. (2025). Top 22 benefits of chatbots for businesses and customers.

https://www.zendesk.com/blog/5-benefits-using-ai-bots-customer-service/