The Impact of Generative AI on IT Professionals’ Work Routines: A

Systematic Literature Review

Davi Maia

, Joao Victor Pereira das Neves, Giovanni Veloso, Guilherme Guerra, Henrique Gomes,

Liliane Carla Oliveira and Simone C. dos Santos

Informatics Center, Federal University of Pernambuco, Av. Jorn. Aníbal Fernandes, Recife, Brazil

Keywords: Generative AI, Software Developer, Work Routine, AI Tool.

Abstract: The rapid evolution of Generative Artificial Intelligence (Generative AI) has had a profound impact on several

industries, including Information Technology (IT). Generative AI is widely recognized for its ability to

automate objective and routine tasks, such as code generation. AI tools, previously restricted to specialists,

are increasingly accessible, expanding their use to a wide range of sectors, including the field of Information

Technology (IT). Considering these impacts is crucial to prepare these professionals for maximize the

potential of new tools in companies and face the challenges that arise with the automation of activities and

processes, as well as with the changes in the skills required. Understanding this context, the present study's

main motivation is to understand how generative AI is reshaping the IT profession, highlighting the

opportunities and challenges that arise with the adoption of these technologies. This research used the

Systematic Literature Review (SLR) in three stages. The analysis of 34 studies made it possible to find some

interesting results. The main activities are code generation, code or script debugging and code documentation.

The main tools are ChatGPT, GitHub Copilot and Tabnine. The main skills developed are prompt formulation,

understanding AI, critical thinking and problem solving.

1 INTRODUCTION

The rapid evolution of Generative Artificial

Intelligence (Generative AI) has had a profound

impact on several industries, including Information

Technology (IT). This research arises from the need

to understand how these technologies are

transforming the work routines of IT professionals, a

group essential for the implementation and

maintenance of innovative technological solutions

(Webb, 2020).

Generative AI is widely recognized for its ability

to automate objective and routine tasks, such as code

generation, the studies reviewed indicate that its

impact goes beyond these basic activities.

The growing production of articles and studies on

generative AI and its impact on society follows the

rapid popularization of these tools among the public.

AI tools, previously restricted to specialists, are

increasingly accessible, expanding their use to a wide

range of sectors, including the field of Information

https://orcid.org/0009-0007-7144-8020

https://orcid.org/0000-0002-7903-9981

Technology (IT). This movement drives debates not

only about the possibilities and opportunities that

generative AI offers, but also about the ethical limits

and implications of its use in the work routines of

various professionals (Fui-noon-Hah, et al., 2023).

These studies reveal that the possibilities for using

AI are expanding rapidly. The tools are becoming

increasingly sophisticated, capable of performing

complex and non-trivial tasks.

This scenario of technological disruption

motivates us to investigate what has already been

impacted and transformed by these innovations,

seeking to understand the changes that are shaping the

present and future of professional routines (IOE,

2024).

Understanding these impacts is crucial to prepare

these professionals for the future of work, maximize

the potential of new tools in companies, and face the

challenges that arise with the automation of activities

and processes, as well as with the changes in the skills

required.

Maia, D., Neves, J. V. P. D., Veloso, G., Guerra, G., Gomes, H., Oliveira, L. C. and Santos, S. C.

The Impact of Generative AI on IT Professionals’ Work Routines: A Systematic Literature Review.

DOI: 10.5220/0013473100003932

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 17th International Conference on Computer Supported Education (CSEDU 2025) - Volume 2, pages 163-173

ISBN: 978-989-758-746-7; ISSN: 2184-5026

163

For example, the article It's like a rubber duck that

talks back: Understanding Generative AI-Assisted

Data Analysis Workflows through a Participatory

Prompting Study [PS9] explores the use of these tools

in more subjective tasks, such as suggesting

evaluation criteria and formulating data analysis

strategies. This demonstrates that generative AI can

act as a partner not only in technical activities, but

also in complex decision-making and less structured

problem-solving.

Understanding this context, the present study's

main motivation is to understand how generative AI

is reshaping the IT profession, highlighting the

opportunities and challenges that arise with the

adoption of these technologies. From this objective,

the following central research question was defined:

"How is Generative Artificial Intelligence

transforming the work routines and skills of IT

professionals?". To answer the central question, the

following secondary questions were defined: Q1)

What are the main types of tasks/activities of IT

professionals that are being automated by Generative

AI? Q2) What are the main tools identified in the

studies? Q3) What changes/skills have been

generated for IT professionals? Q4) What are the

benefits perceived by IT professionals with the

adoption of Generative AI in their activities? Q5)

What are the challenges and barriers faced by IT

professionals when using Generative AI?

This research used the Systematic Literature

Review (SLR) method proposed by (Kitchenham and

Charters, 2007) in three stages: 1) planning the

review; 2) conducting the review and 3) discussion of

the studies. Thus, this paper follows these stages to

present its results.

This paper is organized into six sections. After

this introduction, Sections II describe a brief

conceptual reference in Generative AI. Section III

presents the research methodology. Section IV

presents the results. Section V, presents the

discussions, and Finally, Section VI presents the

conclusions and future perspectives of this research

2 CONTEXT

Generative Artificial Intelligence (Generative AI) is a

sub-area of Artificial Intelligence that uses models

capable of generating data and information from data

previously presented to them. Such content can be in

a variety of formats such as text, images, code, videos

and presentations (Brown et al., 2020).

One of the main techniques within the field of

generative AI is the Generative Adversarial Networks

(GAN), made up of two models: a generator and a

discriminator. The generator model can generate data

based on examples, and the discriminator model is

able to evaluate or distinguish whether the data

generated is real or synthetic. In general, training is

based on data. The generator model receives a larger

sample of random data and depends on feedback from

the discriminator. The discriminator model receives

the largest sample of real data (Goodfellow et al.,

2014).

A more recent technique is the GPT-3 (Generative

Pretrained Transformer 3), which is a model capable

of generating mainly textual data in natural language.

GPT uses a neural network based on transformers that

partitions the text into smaller chunks to generate a

larger number of parameters and more fluid feedback.

Although it mainly generates text, there are tools such

as DALL-E (Ramesh et al., 2021) which can generate

images from descriptions. Despite its benefits and

features, the GPT architecture also has limitations,

such as reproducing prejudices, generating incorrect

data and making it difficult to understand the context

(Vaswani et al., 2017).

3 RESEARCH METHOD

To conduct the research, we used the Systematic

Literature Review methodology and took as a guide

the guidelines proposed by (Kitchenham and

Charters, 2007). This method allowed to identify and

classify relevant studies related to Central Research

Question, as well as to collect and synthesize the

evidence presented in the literature. Figure 1 the steps

followed during the research.

Figure 1: SLR Process.

3.1 Research Planning

In this study, we aim to answer the primary research

question: “How is Generative Artificial Intelligence

CSEDU 2025 - 17th International Conference on Computer Supported Education

164

transforming the work routines and skills of IT

professionals?”

To ensure the answer for this question, we define

a set of secondary questions, to guide the selection

process:

• Q1) What are the main types of

tasks/activities of IT professionals that are

being automated by Generative AI?

• Q2) What are the main tools identified in the

studies?

• Q3) What changes/skills have been

generated for IT professionals?

• Q4) What are the benefits perceived by IT

professionals with the adoption of

Generative AI in their activities?

• Q5) What are the challenges and barriers

faced by IT professionals when using

Generative AI?

After defining the research questions and the

method, we choose the keywords and databases for

primary studies. Four databases were selected: ACM,

IEEE, Scopus, ScienceDirect and Emerald.

From protocol definition we defined keywords to

search string which were used in automatic search, as

shown in Table 1. The keywords are related to

“Generative AI”, “work” and “Software

Development”.

Table 1: Search String.

Search String

(( "Generative AI" OR "LLM" ) AND ( "Software

Engineer" OR "Software Developer" OR

"programmer" ) AND ("work routine" OR "work" OR

"task"

)

AND

(

"AI tool" OR "AI mechanism"

))

3.2 Research Conducting

From search string, automatic searches are conducted

in the selected databases. After automatic search step,

filters are applied to select the relevant and

appropriate studies to research. The first filter applied

on exclusion criteria take account the title, abstract

and keywords of the works. The exclusion criteria

applied in the selected studies, such as:

• E0001 - Articles that are not written in

Portuguese or English

• E0002 - Publications not freely available

• E0003 - Other secondary or tertiary studies

• E0004 - Articles published before 2013

• E0005 - Articles with more than 50 pages

In the filtering, inclusion criteria are used:

• I0001 - Articles related to the research topic

and questions

• I0002 - Full articles

• I0003 - Book chapters, conference papers

The criteria were used in two distinct stages. The first

step used the exclusion criteria. In case of doubt in the

selection, the inclusion criteria are applied reading the

introduction and conclusion of the study.

3.3 Research Reporting

To extract the data, a shared spreadsheet was created,

containing data such as title, year of publication,

source, authors, and important fragments of the

articles' text were extracted and classified. The

columns of fragments from the studies were created

based on the secondary research questions. The

answers are classified according to relevance

information to each question. The quality of answers

was evaluated based on criteria presented in Table 2.

These criteria were scored in a three-point scale: 0 -

Does not meet; 0.5 - Partially meets; 1 - Fully meets.

Table 2: Quality Criteria

ID Description

1Define

methodology

2 Practical a

lication

3 Defined Model

4 Relevant Discussions

5 Challenges, limitations or threats

Due to the small number of articles available for

analysis, we needed to expand our database. To

address this, we used the Snowballing technique,

which consists of identifying relevant references from

the initially selected articles. As an additional

criterion, we performed this search only in articles that

scored the maximum score in the classification (score

5).

After applying this methodology, we were able to

expand our list of articles and selected 10 new articles.

These additional articles were also scored using the

same qualitative criteria at the end of the process to

ensure consistency and reliability in our analysis.

We utilized an AI tool called NotebookML, an

experimental research tool developed by Google, to

assist in answering the secondary questions of our

study. While the tool allows for querying across all

selected articles simultaneously, we encountered

limitations in its performance. Specifically, the tool

occasionally provided answers referencing articles

that were not relevant to the specific question being

asked. To address this issue and improve accuracy, we

The Impact of Generative AI on IT Professionals’ Work Routines: A Systematic Literature Review

165

opted to input and analyze each article individually,

ensuring that the answers directly corresponded to the

content of the chosen article.

3.4 Limitations and Threats

Given the specificity of the topic addressed in this

study, the initial number of articles selected was

relatively low. This limitation can be attributed to the

restricted nature of the subject, which, although

relevant, has a limited number of publications

available within the criteria. Therefore, this research

area (Generative AI), gain focus as in the industry as

in research and academia, with the popularization of

Generative AI tools in society over the last five years.

4 RESULTS

A breadth-first search was conducted using the

Search String in each of the databases, resulting in a

total of 214 articles. After this task, we applied the

exclusion criteria to the titles and abstracts of each

study, resulting in a total of 191 excluded studies and

23 studies accepted.

A second filtering process was conducted, with an

in-depth analysis of the introduction and conclusion

sections of the studies. This step was important to

decide on the inclusion of articles based on the initial

filtering. After this step a total of 24 studies were

accepted.

As mentioned in subsection 3.3, the Snowballing

technique was conducted, resulting in 10 new articles.

Finally, the articles were classified according to

the quality criteria presented in the previous section.

At the end of the process 34 studies were collected.

Figure 2 summarizes the article selection process

according to the PRISMA model.

Figure 2: PRISMA flow chart of Selection Process.

Figure 3 present the studies from different

databases. Scopus source stands the highest number

of relevant works (14/24), followed by Science Direct

(6/24), IEEE (3/24) and Emerald (1/24).

Figure 3: Source of Studies.

Figure 4 presents studies by publication year,

where it is possible to assert that publications starts in

2022 and reaches its highest point in 2024.

Figure 4: Studies in timeline.

During the collection of evidence to answer the

questions defined in the planning, we observed the

following points:

Identification of relevant studies: During the

searches in the databases, we found a significant

number of studies that address the impacts of

generative AI on the work of different professionals.

We selected the articles that: (a) focused on IT

professionals and (b) explicitly mentioned these

professionals.

Specific analyses by area: Some studies collected

carried out analyses focused on specific areas of

technology, such as game development [PS1] and

data analysis [PS9].

Consideration of students: Studies that address the

reality of students who already use generative AI

tools in academic projects were also included [PS2,

PS6, PS7, PS8, SB2]. This allowed us to identify the

CSEDU 2025 - 17th International Conference on Computer Supported Education

166

tools most used by these future professionals and the

changes perceived in the skills they are developing.

4.1 What Are the Main Types of

Tasks/Activities of IT Professionals

that Are Being Automated by

Generative AI?

By analysing the 34 articles selected for the study, 17

types of activities that are being automated in the

work routines of different IT professionals, including

software engineers, programmers (with varying

levels of experience), data analysts and game

developers. Figure 5 presents the activities and the

number of studies which appears.

Figure 5: Automated Activities by Generative AI.

Among the studies reviewed, articles PS9, PS11,

PS12, PS15, SB4 and SB10 stand out for offering a

comprehensive overview of the types of activities

impacted by automation.

The most mentioned task was code generation,

cited in 22 of the 34 articles analysed, representing

approximately 65% of the total, a broad category that

encompasses several specific activities, as evidenced

by the studies. These activities include generation of

boilerplate code [PS11], generate SQL queries

[PS23], auto-complete lines of code and comments

[SB2, SB4], generate code adapted to different

writing styles [SB8], generation of methods and

classes [SB8] and creating code from natural

language comments or prompts [PS10, PS11, PS23,

SB10].

In addition to code generation, data analysis also

encompasses several other activities highlighted in

the studies. These include searching for relevant data

sources, proposing analysis strategies, writing code

for analysis, and suggesting subjective criteria for

evaluating different scenarios. These activities are

well exemplified in the article It's like a rubber duck

that talks back: Understanding Generative AI-

Assisted Data Analysis Workflows through a

Participatory Prompting Study [PS9].

Based on the evidence presented in article PS9, it

was possible to identify that the use of generative AI

goes beyond basic and objective activities, also

extending to more complex and subjective tasks. This

type of action allows AI to function not only as a

technical tool, but also as support in decision-making

and in solving less structured problems.

The Table 3 presents a summary of the types of

activities identified in the studies analysed, along

with their respective references. These types of

activities not only reflect the tasks directly mentioned

in the articles but also serve as broad representations

of other activities that are part of the daily lives of IT

professionals. Thus, each category listed can

encompass a broader set of actions performed in the

routines of software engineers, programmers, data

analysts and game developers, as evidenced in the

reviewed articles.

Table 3: Types of activities by Study.

Types of

tasks/activities

Study

Code generation PS1, PS4, PS6, PS7, PS9, PS10,

PS11, PS12, PS15, PS17, PS18,

PS20, PS23, SB1, SB2, SB4, SB5,

SB6, SB7, SB8, SB9, SB10

Code or script

debugging

PS2, PS7, PS9, PS11, PS12, PS15,

SB4, SB5, SB10

Code documentation PS11, PS12, PS15, PS16, SB2,

SB4, SB8, SB9, SB10

Error monitoring,

detection and

correction

PS4, PS11, PS12, PS15, SB5, SB6,

SB9

Software testing PS4, PS11, PS15, SB4, SB8

Code review PS6, SB5, SB8, SB9

Data analysis PS9, PS15, PS23, SB10

Code explanation PS11, PS20, SB9, SB10

Programming

problem solving

PS2, PS18

Code performance

improvement

PS11, PS12

Requirements

engineering

PS14, PS15

Deploymen

PS15

Graphics

programming

PS1

Q/A (Quality

Assurance)

SB4

Suggesting software

improvements

PS15

DevOps PS14

Cybersecurity PS23

The Impact of Generative AI on IT Professionals’ Work Routines: A Systematic Literature Review

167

4.2 What Are the Main Tools

Identified in the Studies?

A small variety of tools were mentioned in the

articles. Most of them were only mentioned once

among the 34 articles analyzed. However, two tools

stood out and were used more frequently in the works:

ChatGPT, a generative artificial intelligence chatbot

developed by OpenAI, and Github Copilot, an

artificial intelligence tool developed by GitHub in

conjunction with OpenAI, to assist users of integrated

development environments. ChatGPT was used in 11

of the 34 articles, while Copilot was present in 10 of

them. The Figure 6 presents the Generative AI tools

used in works.

Figure 6: Generative AI tools.

Other tools found include Tabnine, an AI coding

assistant designed to be under the control of an

engineering team, Stable Diffusion, which is a deep

learning model for text-to-image transformation,

CodeTutor, which is an LLM-powered assistant

developed by the research team of one of the articles,

which was used by 50 students in order to conclude

the study of that article, Bing Chat, as ChatGPT, is a

chatbot assistant, Twimo, a conceptual framework to

define domain-specific notations, used for the defi-

nition of human driver behaviour and ML-based

services, and Allpy, which provides a library

implementing different active automata learning

algorithms that support the learning of finite state

models of black-box systems.

4.3 What Changes/Skills Have Been

Generated for IT Professionals?

The growing adoption of generative AI is

transforming the profile of skills required of IT

professionals. Traditional skills are expanding to

include a new set of competencies that are essential

for navigating an increasingly AI-driven

environment. To qualitatively analyse the studies, the

macro competencies identified were separated and

grouped according to their occurrence between the

studies as shown in Figure 7.

Figure 7: Generative AI competencies.

Formulating and analyzing prompts have become

essential skills for IT professionals working with

generative AI. This competence involves developing

the skills to create clear and effective instructions that

guide the AI in generating the desired results.

Professionals need to not only understand how to

structure these prompts, but also learn how to

evaluate the AI's responses, adjusting and refining the

requests to improve the accuracy and relevance of the

outputs. This process not only increases efficiency

when interacting with tools such as ChatGPT and

Gemini but also transforms the way professionals

approach complex problems [PS1, SB5, SB7, PS5,

SB3, SB1, PS18].

The ability to think critically and solve complex

problems is amplified by the use of generative AI.

Professionals are being challenged to critically

analyse AI outputs, validating their logic and safety.

This critical approach is crucial to avoiding errors and

biases, promoting more effective use of the

technology [PS2, PS5, SB7, PS16, PS20].

IT professionals need to understand the

capabilities and limitations of AI tools in order to

effectively integrate them into their processes. In-

depth knowledge of generative AI not only increases

effectiveness in implementing technological

solutions but also prepares professionals to innovate

and improve existing systems [PS5, SB5, PS4, PS9,

PS14, PS16, PS23].

The design of intuitive interfaces that

communicate reliability is increasingly important. IT

professionals must be able to create experiences that

consider the needs of end users, ensuring that

interaction with AI systems is transparent and

CSEDU 2025 - 17th International Conference on Computer Supported Education

168

efficient. This competence is essential to ensure speed

in the creation and validation of prototypes and also

broadens the scope of templates and mock-ups [PS9,

PS10, PS11].

Collaboration and communication skills are

amplified in the context of generative AI. IT

professionals must work together with AI systems

and other teams. The ability to articulate complex

ideas clearly and effectively is vital for the successful

implementation of technological projects and can be

best achieved with the intermediation of artificial

intelligence models [PS4, SB1, PS17, PS15, PS16].

With the increased use of AI, awareness of

cybersecurity and AI ethics is paramount. IT

professionals must ensure that solutions respect

security best practices and address ethical issues such

as bias and data privacy. Regulations and laws such

as the LGPD promote caution when adopting

artificial intelligence. IT professionals must pay

attention to security aspects when designing AI

projects so that reliable and reputable systems can be

built [PS5, PS17, SB5].

A continuous learning mindset is essential in a

rapidly evolving field. IT professionals need to be

willing to adapt to new tools and techniques, keeping

up to date with the latest trends in AI. This

willingness to learn and adapt is fundamental to their

professional evolution [PS15, SB2, SB3, PS8, PS16].

Skills in systems development and management

are enhanced by the integration of AI. Professionals

must be able to create and manage complex systems

using AI and machine learning, ensuring efficiency,

effectiveness and innovation in development

processes [PS8, SB4, PS14, PS23].

Finally, creativity is a skill that is stimulated using

generative AI. IT professionals are encouraged to

explore innovative solutions, using AI to generate

ideas and solve complex problems effectively [PS20,

PS18].

4.4 What Are the Perceived Benefits by

IT Professionals with the Adoption

of Generative AI in Their

Activities?

24 of them provide answers or allow the inference of

an answer to the question about the perceived benefits

by IT professionals with the adoption of Generative

AI in their activities. The identified benefits were

grouped into nine main categories, covering aspects

such as learning support, increased productivity, and

improvements in communication and task

automation.

The analysis of the 24 articles that address the

perceived benefits by IT professionals with the

adoption of Generative AI revealed a wide range of

advantages in different areas. The most frequent

benefit was "Support in Learning and Professional

Development," found in 75% of the articles, as shown

in figure 6, followed by "Increased Productivity and

Efficiency" in 66.67%, demonstrating how

Generative AI has the potential to enhance

professionals' capabilities and optimize their time.

Other highlighted benefits include "Improvement in

Code Quality" (45.8%) and "Facilitation of Problem

Solving" (37.5%), showing its value in technical tasks

such as coding and debugging, as shown in Figure 8.

Figure 8: Perceived Benefits.

Among the most mentioned features in Kuhail et

al. (2024) article, "Boilerplate Code Generation" was

the most cited, appearing in 48.5% of the cases. This

highlights how AI automates repetitive tasks,

allowing developers to focus on more complex

activities. Additionally, "Code Explanation" was

noted in 38.4% of cases, with AI acting as a support

for learning and understanding difficult code, helping

professionals speed up their development process.

"Solution Search" (36.4%) and "Error Identification"

(33.3%) were also highlighted as important benefits,

as they optimize information searching and speed up

the debugging process, respectively.

Another important aspect relates to the impact on

productivity. According to Kuhail et al. (2024),

"Faster Coding Speed" was identified in 58.6% of the

cases, with developers reporting that AI accelerates

the process of writing code. Additionally, "More

Effective Code" (27.3%) and "More Concise Code"

(25.3%) were other points mentioned, showing how

AI contributes to producing cleaner code with better

performance. The automation of "Documentation

Writing" (21.2%) and "Test Creation" (18.2%) were

also cited as features that free up time for developers

to focus on more complex tasks.

The Impact of Generative AI on IT Professionals’ Work Routines: A Systematic Literature Review

169

4.5 What Are the Challenges and

Barriers Faced by IT Professionals

when Using Generative AI?

Generative Artificial Intelligence (Generative AI) has

emerged as a powerful tool that can bring numerous

benefits to IT professionals. However, its

implementation faces significant challenges. During

the analysis of the 34 reviewed articles, 19 provided

relevant findings for this question. The Figure 9

presents the main challenges and barriers identified.

Figure 9: Main Challenges by mention.

Workflow disruption (SB1, PS9): Long

suggestions can interrupt programmers' workflow. At

best, these suggestions are immediately discarded,

and at worst, they distract the programmer from their

flow. For instance, upon receiving a 16-line

suggestion and after only four seconds of analysis,

one developer in [SB1] exclaimed: "Oh God, no.

Absolutely not," "Stop it!" and continued

programming as before. On the other hand, many

programmers feel compelled to read the entire code

returned by the AI and noted that reading these long

suggestions often disrupted their flow. Some

perceived interruptions included distraction by the

suggested results and disorientation. One

programmer expressed this frustration: "I was about

to write the code, and I knew what I wanted to write.

But now I’m sitting here, seeing if somehow Copilot

came up with something better than the person who’s

been writing Haskell for five years. I don’t know why

I am giving it the time of day." [SB1] These

distractions cause some programmers to abandon the

tool altogether.

Difficulty in understanding, validating, and

debugging (SB1, SB4, SB8, SB10, PS18): Some

developers report difficulties in understanding,

validating, and debugging the code generated by AI

tools. The lack of immediate familiarity with the

suggested code makes error identification more time-

consuming and complex. As one developer

commented: “I don’t see the error immediately, and

unfortunately, because this is generated, I don’t

understand it as well as I feel like I would’ve if I had

written it. I find reading code that I didn’t write to be

a lot more difficult than reading code that I did write,

so if there’s any chance that Copilot is going to get it

wrong, I’d rather just get it wrong myself because at

least that way I understand what’s going on much

better.” [SB1] Professionals claim that since they did

not write the code, their understanding of errors is

impaired, making debugging more challenging. They

observe that reading and interpreting code generated

by others is significantly more difficult than working

with code they developed themselves. “Participants

reported spending less time on Stack Overflow but

now have less understanding of how or why the code

works.” [SB8] In academic contexts, this difficulty is

even more evident. The superficial use of these tools,

merely to get answers, can prevent students from

developing a complete understanding of

programming principles. Copying and pasting code

without understanding the logic behind it can be

detrimental in the long run. [SB10] In this sense,

some developers prefer to make their own mistakes

while writing code, as it provides a clearer and deeper

understanding of what is happening in the system,

facilitating correction and learning.

Lack of trust and control challenges (SB1, SB2,

SB3, SB4, SB6, SB7, PS2, PS4, PS6, PS7, PS9, PS10,

PS11, PS15, PS17, PS18): Many programmers report

not fully trusting the code generated by AI tools. As

one developer mentioned: "It’s not official

documentation, it’s something that needs my

examination...if it works, it works." [SB1]. In [SB2],

it was identified that with Copilot, some of the

suggestions are often wrong, include unnecessary

elements, or are mainly variations on a theme. As

observed by a participant: "Copilot most often does

not understand our instructions to fix or improve the

code it generated unless we formulate them in a very

specific way." This problem was also identified with

ChatGPT [SB3], which, although it can provide

correct answers to many questions related to bug

fixing, the overall accuracy rate is still relatively low.

Developers also face challenges in controlling AI

tools, as reported in [SB4]: "The most important

reasons why developers do not use these tools are

because these tools do not output code that addresses

certain functional or non-functional requirements and

because developers have trouble controlling the tool

to generate the desired output."

Data security and privacy (SB3, SB8, PS6, PS15):

IT professionals express concerns about the security

and privacy of data used in training large language

models (LLMs), which can be a barrier to adoption.

When it comes to adopting ChatGPT for bug fixing,

data security is a major concern for many developers.

The study [SB3] highlights that many developers

cited concerns about data confidentiality as the reason

CSEDU 2025 - 17th International Conference on Computer Supported Education

170

they would not use ChatGPT for bug fixing. They

were worried that the system might leak sensitive

information, putting their companies and clients at

risk. Moreover, they do not want their inputs and

outputs to be stored by the system and potentially

shared with other users later. Similarly, in the study

[SB8], Copilot users identified some challenges,

including the risk of revealing secrets like API keys

and passwords, suggesting inappropriate text, and

failing to write “defensive code,” such as checking

null pointers. The study [PS6] also presents that in

terms of code quality, AI tools can generate code that

is not robust and may have security vulnerabilities.

Difficulty in communicating intentions and

preferences to the AI (PS1, PS9, PS10, PS17):

Developers face difficulties in clearly communicating

their intentions and preferences to code-generation

tools, which can result in unsatisfactory outputs. In

study [PS9], it was observed that "Part of the

challenge was in fully articulating their need.

Participants had trouble ‘wording it in the right way

that the AI understands [...] writing [what is in your

head] down is the hard part.’" Developers found it

difficult to express their intentions in a way that the

AI could correctly understand what they wanted. This

challenge was also noted in [PS17], which details

how developers using Copilot struggle to generate the

expected results. Often, they need to invest

considerable effort in crafting strategies to design

prompts and debug the model's inputs. This

emphasizes the importance of proper prompt

construction for the successful use of these tools.

5 DISCUSSIONS

The findings of systematic literature review reveal

that generative AI is significantly transforming the

work routines and required skill sets of IT

professionals. The analysis of 34 studies shows that

generative AI tools, such as GitHub Copilot and

ChatGPT, are automating a variety of tasks, primarily

in code generation and data analysis. These tools

provide substantial advantages in terms of increased

productivity, reduced time spent on repetitive tasks,

and enhanced learning opportunities for both

experienced professionals and students.

However, the adoption of these tools is not

without challenges. One of the major issues reported

by the studies is the disruption of workflow caused by

overly complex or irrelevant suggestions from AI

tools, which can sometimes distract or slow down

developers. Additionally, there are concerns

regarding the accuracy of AI-generated code, with

professionals expressing difficulties in

understanding, validating, and debugging the

suggestions provided by these tools. These challenges

highlight the need for further development of AI

systems to improve the relevance and reliability of

their outputs.

Another significant barrier identified in the

literature is the trust and control challenge. Many

developers remain sceptical of AI-generated outputs,

particularly due to the lack of transparency in how the

AI reaches its conclusions. Moreover, data privacy

and security concerns present a substantial barrier,

especially when using tools that rely on large datasets,

some of which may contain sensitive information.

Despite these challenges, the introduction of

generative AI tools has led to the emergence of new

skill sets for IT professionals. Prompt formulation,

critical thinking, and AI tool proficiency have

become increasingly important. The ability to

collaborate effectively with AI systems and

continuously learn and adapt to evolving technologies

has also been highlighted as critical to success in an

AI-driven environment.

6 CONCLUSIONS

The adoption of generative AI within the IT sector

offers both opportunities and challenges. On one

hand, these tools enable the automation of routine

tasks, such as code generation, freeing professionals

to focus on more complex and creative aspects of

their work. On the other hand, issues such as trust,

workflow disruptions, and data security remain

substantial barriers to widespread adoption.

The review demonstrates that while AI tools can

enhance productivity and efficiency, their use must be

carefully managed to avoid over-reliance, which can

hinder deeper understanding and development of core

programming skills, particularly in educational

contexts.

Despite the benefits presented, the authors have

encountered some challenges in the conduction

process, related to the low number of studies and the

recent nature of the topic. These challenges could be

addressed with snowballing techniques, but in a

general context, it is a call for more research in the

field of Education and Generative AI.

To fully harness the potential of generative AI,

future efforts should focus on addressing these

challenges through the development of more

transparent, accurate, and secure AI systems.

Additionally, ongoing research is necessary to

evaluate the long-term effects of AI adoption on

professional development and to establish best

practices for integrating these technologies into daily

workflows.

The Impact of Generative AI on IT Professionals’ Work Routines: A Systematic Literature Review

171

REFERENCES

Webb, M. (2019). The impact of artificial intelligence on

the labor market. Available at SSRN 3482150.

Fui-Hoon Nah, F., Zheng, R., Cai, J., Siau, K., & Chen, L.

(2023). Generative AI and ChatGPT: Applications,

challenges, and AI-human collaboration. Journal of

Information Technology Case and Application

Research, 25(3), 277-304.

International Organization of Employers (IOE). (2024).

The Impact of AI on Work and Employment.

Brown, T., Mann, B., Ryder, N., Subbiah, M., Kaplan, J.

D., Dhariwal, P., ... & Amodei, D. (2020). Language

models are few-shot learners. Advances in neural

information processing systems, 33, 1877-1901.

Goodfellow, I., Pouget-Abadie, J., Mirza, M., Xu, B.,

Warde-Farley, D., Ozair, S., ... & Bengio, Y. (2014).

Generative adversarial nets. Advances in neural

information processing systems, 27.

Ramesh, A., Pavlov, M., Goh, G., Gray, S., Voss, C.,

Radford, A., Chen, M. & Sutskever, I.. (2021).

Zero-Shot Text-to-Image Generation. Proceedings of

the 38th International Conference on Machine

Learning in Proceedings of Machine Learning

Research 139:8821-8831 Available from

https://proceedings.mlr.press/v139/ramesh21a.html.

Vaswani, A. (2017). Attention is all you need. Advances in

Neural Information Processing Systems.

Kitchenham, B. and Charters, S. (2007) Guidelines for

Performing Systematic Literature Reviews in Software

Engineering, Technical Report EBSE 2007-001, Keele

University and Durham University Joint Report.

APPENDIX A

Selected Studies

[PS1] Colado, I. J. P., Colado, V. M. P., Morata, A. C.,

Píriz, R. S. C., & Manjón, B. F. (2023, October). Using

new AI-driven techniques to ease serious games

authoring. In 2023 IEEE Frontiers in Education

Conference (FIE) (pp. 1-9). IEEE.

[PS2] Yilmaz, R., & Yilmaz, F. G. K. (2023). Augmented

intelligence in programming learning: Examining

student views on the use of ChatGPT for programming

learning. Computers in Human Behavior: Artificial

Humans, 1(2), 100005.

[PS3] Ferdowsi, K., Williams, J., Drosos, I., Gordon, A. D.,

Negreanu, C., Polikarpova, N., ... & Zorn, B. (2023,

October). ColDeco: An end user spreadsheet inspection

tool for AI-generated code. In 2023 IEEE Symposium

on Visual Languages and Human-Centric Computing

(VL/HCC) (pp. 82-91). IEEE.

[PS4] Javaid, M., Haleem, A., & Singh, R. P. (2023). A

study on ChatGPT for Industry 4.0: Background,

potentials, challenges, and eventualities. Journal of

Economy and Technology, 1, 127-143.

[PS5] Peres, R., Schreier, M., Schweidel, D., & Sorescu, A.

(2023). On ChatGPT and beyond: How generative

artificial intelligence may affect research, teaching, and

practice. International Journal of Research in

Marketing, 40(2), 269-275.

[PS6] Styve, A., Virkki, O. T., & Naeem, U. (2024, May).

Developing critical thinking practices interwoven with

generative AI usage in an introductory programming

course. In 2024 IEEE Global Engineering Education

Conference (EDUCON) (pp. 01-08). IEEE.

[PS7] Programming Boguslawski, S., Deer, R., & Dawson,

M. G. (2024). Programming education and learner

motivation in the age of generative AI: student and

educator perspectives. Information and Learning

Sciences.and learner motivation in the age of generative

AI: student and educator perspectives

[PS8] Lyu, W., Wang, Y., Chung, T., Sun, Y., & Zhang, Y.

(2024, July). Evaluating the effectiveness of llms in

introductory computer science education: A semester-

long field study. In Proceedings of the Eleventh ACM

Conference on Learning@ Scale (pp. 63-74).

[PS9] Drosos, I., Sarkar, A., Xu, X., Negreanu, C., Rintel,

S., & Tankelevitch, L. (2024, June). “It’s like a rubber

duck that talks back”: Understanding Generative AI-

Assisted Data Analysis Workflows through a

Participatory Prompting Study. In Proceedings of the

3rd Annual Meeting of the Symposium on Human-

Computer Interaction for Work (pp. 1-21).

[PS10] Wang, R., Cheng, R., Ford, D., & Zimmermann, T.

(2024, June). Investigating and designing for trust in ai-

powered code generation tools. In The 2024 ACM

Conference on Fairness, Accountability, and

Transparency (pp. 1475-1493).

[PS11] Kuhail, M. A., Mathew, S. S., Khalil, A.,

Berengueres, J., & Shah, S. J. H. (2024). “Will I be

replaced?” Assessing ChatGPT's effect on software

development and programmer perceptions of AI tools.

Science of Computer Programming, 235, 103111.

[PS12] Önden, A., Kara, K., Önden, İ., Yalçın, G. C., Simic,

V., & Pamucar, D. (2024). Exploring the adoption of

the metaverse and chat generative pre-trained

transformer: A single-valued neutrosophic Dombi

Bonferroni-based method for the selection of software

development strategies. Engineering Applications of

Artificial Intelligence, 133, 108378.

[PS13] Haleem, A., Javaid, M., & Singh, R. P. (2024).

Exploring the competence of ChatGPT for customer

and patient service management. Intelligent Pharmacy.

[PS14] Eramo, R., Said, B., Oriol, M., Bruneliere, H., &

Morales, S. (2024). An architecture for model-based

and intelligent automation in DevOps. Journal of

Systems and Software, 217, 112180.

[PS15] Russo, D. (2024). Navigating the complexity of

generative ai adoption in software engineering. ACM

Transactions on Software Engineering and

Methodology.

[PS16] Feldman, M. Q., & Anderson, C. J. (2024, June).

Non-Expert Programmers in the Generative AI Future.

In Proceedings of the 3rd Annual Meeting of the

Symposium on Human-Computer Interaction for Work

(pp. 1-19).

CSEDU 2025 - 17th International Conference on Computer Supported Education

172

[PS17] Cheng, R., Wang, R., Zimmermann, T., & Ford, D.

(2024). “It would work for me too”: How Online

Communities Shape Software Developers’ Trust in AI-

Powered Code Generation Tools. ACM Transactions

on Interactive Intelligent Systems, 14(2), 1-39.

[PS18] Ramaul, L., Ritala, P., & Ruokonen, M. (2024).

Creational and conversational AI affordances: How the

new breed of chatbots are revolutionizing the

knowledge industries. Business Horizons.

[PS19] Climent, R. C., Haftor, D. M., & Staniewski, M. W.

(2024). AI-enabled business models for competitive

advantage. Journal of Innovation & Knowledge, 9(3),

100532.

[PS20] Messer, U. (2024). Co-creating art with generative

artificial intelligence: Implications for artworks and

artists. Computers in Human Behavior: Artificial

Humans, 2(1), 100056.

[PS21] Varma, A., Pereira, V., & Patel, P. (2024). Artificial

intelligence and performance management.

Organizational Dynamics, 53(1), 101037.

[PS22] Carroll, A. J., & Borycz, J. (2024). Integrating large

language models and generative artificial intelligence

tools into information literacy instruction. The Journal

of Academic Librarianship, 50(4), 102899.

[PS23] Handler, A., Larsen, K. R., & Hackathorn, R.

(2024). Large language models present new questions

for decision support. International Journal of

Information Management, 79, 102811.

[PS24] Ivcevic, Z., & Grandinetti, M. (2024). Artificial

intelligence as a tool for creativity. Journal of

Creativity, 34(2), 100079.

[SB1] Barke, S., James, M. B., & Polikarpova, N. (2023).

Grounded copilot: How programmers interact with

code-generating models. Proceedings of the ACM on

Programming Languages, 7(OOPSLA1), 85-111.

[SB2] Wermelinger, M. (2023, March). Using github

copilot to solve simple programming problems. In

Proceedings of the 54th ACM Technical Symposium on

Computer Science Education V. 1 (pp. 172-178).

[SB3] Ge, H., & Wu, Y. (2023). An Empirical Study of

Adoption of ChatGPT for Bug Fixing among

Professional Developers. Innovation & Technology

Advances, 1(1), 21-29.

[SB4] Liang, J. T., Yang, C., & Myers, B. A. (2024,

February). A large-scale survey on the usability of ai

programming assistants: Successes and challenges. In

Proceedings of the 46th IEEE/ACM International

Conference on Software Engineering (pp. 1-13).

[SB5] Atkinson, C. F. (2023). ChatGPT and computational-

based research: benefits, drawbacks, and machine

learning applications. Discover Artificial Intelligence,

3(1), 42.

[SB6] Jalil, S., Rafi, S., LaToza, T. D., Moran, K., & Lam,

W. (2023, April). Chatgpt and software testing

education: Promises & perils. In 2023 IEEE

international conference on software testing,

verification and validation workshops (ICSTW) (pp.

4130-4137). IEEE.

[SB7] Dakhel, A. M., Majdinasab, V., Nikanjam, A.,

Khomh, F., Desmarais, M. C., & Jiang, Z. M. J. (2023).

Github copilot ai pair programmer: Asset or liability?.

Journal of Systems and Software, 203, 111734.

[SB8] Bird, C., Ford, D., Zimmermann, T., Forsgren, N.,

Kalliamvakou, E., Lowdermilk, T., & Gazit, I. (2022).

Taking Flight with Copilot: Early insights and

opportunities of AI-powered pair-programming tools.

Queue, 20(6), 35-57.

[SB9] Biswas, S. (2023). Role of ChatGPT in Computer

Programming. Mesopotamian Journal of Computer

Science, 2023, 9-15.

[SB10] Silva, C. A. G. D., Ramos, F. N., de Moraes, R. V.,

& Santos, E. L. D. (2024). ChatGPT: Challenges and

benefits in software programming for higher education.

Sustainability, 16(3), 1245.

The Impact of Generative AI on IT Professionals’ Work Routines: A Systematic Literature Review

173