Leveraging Artificial Intelligence in Project Management: A Systematic Review of Applications, Challenges, and Future Directions

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Review articles serve an important role in making it easier for researchers to understand the current state of knowledge in a particular field. I enjoy reading such articles, and expect such an article will identify patterns and gaps in the literature, and allow a critical evaluation of the field.

 

I am going to read through the Abstract very carefully right now, as it is the part of the paper that will be read more than any other part. Ok…. first sentence…. “AI is defined as the study of intelligent computational agents [1,2] with a significant impact in Project Management..”  Really?  Ok… let me look at the references [1,2] – what?? My background/work is in neuroscience and in artificial intelligence. So right away at the start of the Abstract I am wondering how much this article is really related to AI or computers. Why not start by saying what the paper is really about, i.e., reviewing AI methodologies in project management. Ok… next sentence…. What is reference #3 – no volume numbers, etc. – where do I find this reference?

 

(Also, the way the first sentence is written provides a very limited definition of AI because AI encompasses far more than project management applications. I don’t know if you really need to define artificial intelligence in your Abstract as all of readers of this journal will know the term, but if you are going to do so, please use a more traditional reference, e.g., the Russell and Norvig reference or many equivalents.)

 

I apologize for being so critical in the first two sentences of the Abstract, but the reality is that I read every paper like this—the paper, the ideas, the references have to make sense. If you take away confidence from the reader right at the start, then the reader starts wondering if it is really worth making so much effort to scrutinize the paper so carefully, but rather the reader may very well read through the paper more superficially, as I will do now.

 

The Introduction seems to be better written. The first sentence sort of starts off with the subject this paper is about…. Well I think it does, although I’m not really sure. Line 54 is a more realistic consideration of what you consider AI to be, and I am ok with this sentence.

 

Ok… finally at line 82 I know what this paper is about: “This research offers a systematic literature review to thoroughly examine the role of 82 AI in PM. By analyzing the utilization of AI methodologies on critical project success fac- 83 tors (CSFs), such as cost estimation, duration forecasting, and risk assessment, this study 84 focuses on enlighten researchers, stakeholders, and practitioners about the potentials of 85 AI in advancing PM techniques.”

 

 Line 121 – Thank you for showing details how you approached your PRISMA work. Figure 1 is a good start but please make more of an effort on the captions—tell the reader in the captions what you are doing here.

 

 Ok… I am now at Section 3 and at this point I want to see what you found. Line 189 – ACO algorithm – I’ve actually done work in this area so I know what it means, but I don’t think most readers will—this is a computers journal, not an OR journal. It is in areas like this that a reference (and/or a sentence explaining what it is) is most useful. For example, Line 404 you describe RIDOR – thank you as I was not familiar with this before reading it, so as a reader it useful understanding what I am reading. (A reference would be nice for concepts which are not well known to the readership.)

 

Line 452 paragraph – much too long. Consider summarizing better the ideas you want to say and break up the paragraph. Same thing for following paragraphs. At this point, as the reader, I want to see a bit more of a theme, I want to be guided in the reading of all this information rather than to have it all splashed in front of me.

 

 Line 577 where you start to analyze a bit your literature review and introduce Figure 3 – perhaps this should be a new section or subsection?

 

I am now at line 619 where Section 4 starts. At this point I feel like the previous section was too much of a whirlwind with the analysis just shoved in at the end. I’d really like to see Figures 3, 4, 5, 6 and 7 in a coherent section or subsection of their own. Ok… the Conclusion. This is a very important part of the paper. The section does identify gaps and challenges, but it does not really provide an in-depth discussion of why these gaps exist or approaches towards them. My largest criticism is that this is supposed to be an academic paper, not a consumer review of appliances or autos, etc. As an academic review paper, this paper should really connect its conclusions to broader theoretical frameworks or paradigms in AI and PM. This section largely focuses on practical gaps without adequately discussing theoretical implications or advancements.

 

 Please accept my criticisms with an open mind to producing a better paper for the readers of this journal. (Typed without reading – please excuse any small typos.)

Author Response

Comments 1:  I am going to read through the Abstract very carefully right now, as it is the part of the paper that will be read more than any other part. Ok…. first sentence…. “AI is defined as the study of intelligent computational agents [1,2] with a significant impact in Project Management..”  Really?  Ok… let me look at the references [1,2] – what?? My background/work is in neuroscience and in artificial intelligence. So right away at the start of the Abstract I am wondering how much this article is really related to AI or computers. Why not start by saying what the paper is really about, i.e., reviewing AI methodologies in project management. Ok… next sentence…. What is reference #3 – no volume numbers, etc. – where do I find this reference?Response 1: Thank you for pointing these out. We proceeded in changing the beginning of the Abstract and removing the reference #3 - page 1, abstract, line 12.

Comments 2:  Line 121 – Thank you for showing details how you approached your PRISMA work. Figure 1 is a good start but please make more of an effort on the captions—tell the reader in the captions what you are doing here.

Response 2: We procceeded in adding more details in the caption of Figure 1 - page 4, methodology, line 130.

Comments 3: Line 452 paragraph – much too long. Consider summarizing better the ideas you want to say and break up the paragraph. Same thing for following paragraphs. At this point, as the reader, I want to see a bit more of a theme, I want to be guided in the reading of all this information rather than to have it all splashed in front of me. ine 577 where you start to analyze a bit your literature review and introduce Figure 3 – perhaps this should be a new section or subsection?

Response 3: We decided to follow your guidance, and the comments of the editor, who sugested to create a classification analysis and also an evolution path in order to make section 3 less flat. We changed the structure of section 3, by adding subparagraphs, classificatoin analysis and an evolution path - (pages 06-18, Section 3, lines 192-701).

Comments 4: I am now at line 619 where Section 4 starts. At this point I feel like the previous section was too much of a whirlwind with the analysis just shoved in at the end. I’d really like to see Figures 3, 4, 5, 6 and 7 in a coherent section or subsection of their own. Ok… the Conclusion. This is a very important part of the paper. The section does identify gaps and challenges, but it does not really provide an in-depth discussion of why these gaps exist or approaches towards them. My largest criticism is that this is supposed to be an academic paper, not a consumer review of appliances or autos, etc. As an academic review paper, this paper should really connect its conclusions to broader theoretical frameworks or paradigms in AI and PM. This section largely focuses on practical gaps without adequately discussing theoretical implications or advancements.

Response 4: We agree with your point and we tried to expand the conclusion section in order to create a more understandable context - (page 19, Section 4, line 712)

 

Reviewer 2 Report

Comments and Suggestions for Authors

This systematic review explores the integration of Artificial Intelligence in Project Management with the focusing on its applications, challenges, and future directions. It synthesizes findings from 98 studies published between 2011 and 2024 by highlighting the role of AI in optimizing project success through cost estimation, duration forecasting, and risk assessment. The paper covers diverse range of AI techniques such as Machine Learning, Deep Learning, Natural Language Processing, and hybrid models trying to cover a broad overview of the field. However, I would like to get some takeways from the review by understanding better the mentioned limitations. In addition, a good review paper should mention ways of tackling these issues or limitations (which are usually mentioned as persepectives in the reviewed papers). So, according to this, I summarize some remarks and suggestions:

 

- Definition of basic notions such as machine learning, deep learning, neural network, fuzzy logic should be removed. They are well known and can be found anywhere.

- The review includes extensive details but lacks thematic grouping. The third section is too long and not well structured, which could lead to reader disengagement.

- The review mentions that existing models fail to account for dynamic project conditions, but it does not extensively analyze the reasons behind this limitation. A deeper critique or suggestions for integrating dynamic factors into AI models could enhance the discussion.

 

- The authors note that many AI models and methodologies in PM are not validated with real-world projects, which limits practical applicability. This is acknowledged as a gap, but the paper could further emphasize the implications of this issue and recommend specific actions for future researchers.

 

 

- The review highlights that most studies focus on the planning and execution stages of projects, with minimal attention to project closure and post-project evaluation. So, the authors could propose actionable solutions or frameworks to guide future studies on underrepresented phases.

 

- The paper makes broad claims about the effectiveness of AI in PM but does not delve into the potential biases or limitations of AI models, such as the overfitting of machine learning models or the challenges in generalizing results across industries.

 

 

- A significant portion of the reviewed studies focuses on the construction sector. So to not limit the generalizability of findings, I recommend to include applications in other industries.

 

- What are the the pros and cons of using hybrid models? The authors can mentions the trade-offs higher computational demands versus improved accuracy, and propose ways to mitigate these challenges.

- The fonts of some figures are not clear and should be improved. Also, their explanation can be enhanced, particularly Figures 5, 6, and 7.

- The conclusion is somewhat repetitive and could provide more actionable insights rather than reiterating points already discussed.

 

Author Response

Comments 1: Definition of basic notions such as machine learning, deep learning, neural network, fuzzy logic should be removed. They are well known and can be found anywhere.

Response 1: Thank you for your comments. We agree with you. We procceeded in deleting this section of information.

Comments 2: The review includes extensive details but lacks thematic grouping. The third section is too long and not well structured, which could lead to reader disengagement.

Respnse 2: Following your guidance, we changed the structure of Section 3. The editore suggested to add a classification analysis and an evolution path, so the section 3 is devided in subparagraphs. - (page 6-18, Section 3, lines 192-693)

Coments 3: The review mentions that existing models fail to account for dynamic project conditions, but it does not extensively analyze the reasons behind this limitation. A deeper critique or suggestions for integrating dynamic factors into AI models could enhance the discussion. -The authors note that many AI models and methodologies in PM are not validated with real-world projects, which limits practical applicability. This is acknowledged as a gap, but the paper could further emphasize the implications of this issue and recommend specific actions for future researchers.- The review highlights that most studies focus on the planning and execution stages of projects, with minimal attention to project closure and post-project evaluation. So, the authors could propose actionable solutions or frameworks to guide future studies on underrepresented phases. The conclusion is somewhat repetitive and could provide more actionable insights rather than reiterating points already discussed.

Response 3: Thank you. We add a more detailed conclusion, trying to follow your suggestions - page 19, line 712

Comments 4;  A significant portion of the reviewed studies focuses on the construction sector. So to not limit the generalizability of findings, I recommend to include applications in other industries

Response 4: during the analysis of the articles we found that the articles referring to the construction works. However, the analysis followed the Prisma methodology and it is quite difficult to change the entire structure and steps of the analysis.

Comments 5: The review highlights that most studies focus on the planning and execution stages of projects, with minimal attention to project closure and post-project evaluation. So, the authors could propose actionable solutions or frameworks to guide future studies on underrepresented phases.- The paper makes broad claims about the effectiveness of AI in PM but does not delve into the potential biases or limitations of AI models, such as the overfitting of machine learning models or the challenges in generalizing results across industries.

Response 5: Section 4 now contains a more detailed gap section adn future research section - page 19, lines 727 - 757

Comments 6: - The fonts of some figures are not clear and should be improved. Also, their explanation can be enhanced, particularly Figures 5, 6, and 7.

Response 6: Fonts are changed now. We added some figures for Figure 3, in order someone to "zoom" in, as the decision tree is larger than the others. - page 17, line 672

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

All of my remarks have been addressed in the revised version of the paper, either partly or fully. The quality of the paper is improved either in terms of structure, content depth, and clarity. I have only one minor remark and another time-consuming one:

- The new subsection 3.10 about evolution path is not well written. A figure showing the time lines can help to improve it. In addition, the authors can add more description and place it among the first subsections.

- What I would have liked to see in the paper, and forgot to mention in my previous review, is an exploration of the data used in the reviewed studies. This should include an assessment of data quality, how closely it resembles complex real-world data, and details on data collection methods and sources. Since the methods discussed are data-driven, this information is crucial for readers to understand the reliability and applicability of the findings and it will also help them to select or find data. Therefore, I recommend that the authors add a subsection to discuss these aspects.

Author Response

Comments 1:  The new subsection 3.10 about evolution path is not well written. A figure showing the time lines can help to improve it. In addition, the authors can add more description and place it among the first subsections.

Response 1: Thank you for your feedback. We proceeded in adding a figure to explain graphically the evolution path. The reason why we did not add it among the first subsections it is because the data for the evolution path are extracted from the literature review above (page 15, line 630)

Comments 2: What I would have liked to see in the paper, and forgot to mention in my previous review, is an exploration of the data used in the reviewed studies. This should include an assessment of data quality, how closely it resembles complex real-world data, and details on data collection methods and sources. Since the methods discussed are data-driven, this information is crucial for readers to understand the reliability and applicability of the findings and it will also help them to select or find data. Therefore, I recommend that the authors add a subsection to discuss these aspects.

Response 2: We add a new subsection in Section 3 (3.11). We tried to include data from severous articles in order to comment on the data quality (page 19, line 703).

Thank you for your feedback. We really hope that we addressed your remarks and we created an article based on the journal's expectations.