Journal Publication of International Research for Engineering and Management (JOIREM)

Volume: 10 Issue: 01 | Jan-2025

AI In Modern Software Development: Current Practices, Challenges and Future

Possibilities .
Sunil Medepalli
IBM
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Artificial Intelligence is fast emerging as an Things, and others have significantly influenced
inseparable part of software engineering. Its impact on software engineering in various ways. However,
modern software development is significantly high.
still, the role of experimental AI is more vital. It has
Though writing code or to be more precise, generating
code through AI systems and tools facilitates the ushered the software industry into a new phase of
development of a wide variety of products, there are “anything possible” where any ideas could easily be
ethical considerations, risks, and limitations that turned into a solution.
ultimately affect the overall quality of the software. This
research paper highlights how AI is used in modern
software design, development, testing, security, and Artificial Intelligence is now widely used by
management processes and they are flawlessly corporate, and small and big companies in
streamlined for enhanced productivity and quality. The simplifying their software development processes.
research describes how AI is important in software
engineering as firms without employing AI in SDLC are
It enables developers to overcome a plethora of
susceptible to underperforming. It provides insights into, problems in core areas of coding, code refinement,
what it could help with in terms of developing modern automation, code optimization, debugging,
and scalable applications, and the ethics involved in AI- deployment, and project management (Barenkamp
driven software engineering. It also highlights
limitations underlying the applications of AI in the past,
et al., 2020). ML, an AI’s subset helps analyze
present, and future. large sets of data and information from previous
projects. The predictive analytics guide teams in
making well-informed and insightful decisions to
1.INTRODUCTION
further improve their current or ongoing projects.
Software Development is rapidly changing, NLP (Natural Language Processing) is a key to
growing in strength with the power of new building more meaningful interactions. A wide
technologies. Earlier, it was hard for anyone to range of code development tools streamline team
imagine a bot could be capable enough of collaboration and improve efficiency and
simplifying human tasks (Hernández-Ledesma et transparency in project management (Amershi et
al., 2017). No doubt, the modern world is software- al., 2020). Apart from this, Marar (2024) explains
obsessed, running various types of software and how AI as a tool is more powerful when it comes to
programs that make work more efficient and CI/CD (continuous integration and continuous
reliable (Farley, D., 2021). Seemann (2021) states development). Shneiderman (2020) describes how
that the primary goal of software engineering is to identifying small and big changes in code at any
develop a more reliable, efficient and user-friendly stage like designing, development, testing or
system. Whether it is a small application or a big security has always been more crucial and how AI
application, it must address the goals. Different makes it easier with minimal or even without
technologies from time to time have successfully human intervention. Adopting AI strategies,
been implemented, and have greatly influenced the developers can take on agile methodologies to
way the imaginations are programmed (Pantiuchina swiftly respond to market requirements and deliver
et al., 2017). Indeed, technologies such as cloud more intuitive and robust products.
computing, data science, blockchain, the Internet of
Software Engineering Evolution
© 2025, JOIREM |www.joirem.com| Page 1
 Journal Publication of International Research for Engineering and Management (JOIREM)
Volume: 10 Issue: 01 | Jan-2025

beneficial in terms of developing smart solutions

Software history is certainly amazing from the first within less time. The use of AI in software
program designed by Ada Lovelace in the mid-19th engineering increased over the past two decades. It
century to the birth of ENIAC (Electronic is now a technique commonly employed in
Numerical Integrator and Computer) programming designing, building, and testing various software
in the 1940s. Though it moved through various components with ease. This in turn enables a
evolutionary phases, it reached new levels of smooth development process. But, still, there is also
possibilities with the introduction of AI in the mid- a supposed gap between the research in AI
20th century. Early AI-based systems were good at applications and the actual implementation of AI in
performing a wide range of tasks in problem- the software programming processes.
solving and logical reasoning (Philipson, 2004). AI
techniques continued to evolve since then. The use Generating, Analyzing, and Revamping Code:
of symbolic reasoning and heuristic search became Synching AI with Human Intelligence for Better
a little practical. ELIZA an AI program, a natural Results
language processing program could be an example
of its growing influence then. Mijwel (2015) Creating the required code as per the project goals
explained how with the development of programs is one big thing for all developers. Ideas are turned
that could show human-like intelligence in the into real applications only with proper coding. If it
1980s (an example of it could be MYCIN) new goes wrong, the entire application goes wrong. But,
milestones in AI in software engineering were writing the code is now a thing of the past. We use
reached. Currently, popular techniques such as the term “code generation” as these are AI tools
neural networks facilitate the process of software that help with creating the right code snippets. Now
testing and development process optimization. The tools come with the power of AI, trained on a large
real impact of AI surfaced when AI was preferred set of programs, code structures, architecture,
as a technology to generate code for small and big designs, and patterns which also seem to have the
software programs. Bot-driven, AI-powered coding, potential to solve even complex problems (Kulkarni
debugging, application security, testing, et al., 2017). Human-like intelligence was limited in
monitoring, and management are now common. the past but AI technology made it all possible for
But, yes, all these have certain limitations. Some systems and machines to think and act like humans
problems and challenges exist and they largely and generate code at speed. Analyzing the context,
impact the development of worthier, more user- the programs now generate code snippets that
friendly, and business-friendly applications (Vaidya depict the developers’ and the stakeholders’
et al., 2023). thoughts.

SDLC and Automation Possibilities No doubt, it is the power of NLP (natural language
processing) that has made this task much simpler
Software engineering involves a large number of since it allows the users to drape their thoughts with
tasks. De Silva et al. (2022) described the the real code, as they could provide instructions in
complexities involved in different phases in the plain language to get the right or the preferred
Software Development Lifecycle (SDLC) and how output. By describing code requirements in natural
each of these phases will have small goals. Liu, et language, coders can simplify their task of coding.
al., (2020) stress on importance of AI applications For instance, developing a more secured and
or approaches which are known to be more scalable REST API endpoint to extract, define, and

© 2025, JOIREM |www.joirem.com| Page 2

 Journal Publication of International Research for Engineering and Management (JOIREM)
Volume: 10 Issue: 01 | Jan-2025

refine the user data is now made simple with accuracy and correctness of such code snippets are
instructions given to the AI-code generation tool. also one of the biggest issues.
Similarly, getting suggestions, and
recommendations on the next block of code or to AI's capabilities to identify and analyze patterns
put it simply, code completion support increases and deliver more powerful predictions enable a
productivity greatly. For instance, systems trained team of developers to overcome barriers in
in Python, Java, and other coding languages development, reduce or overcome inconsistencies
provide developers with suggestions or what could or errors, and ultimately improve deployment and
be the next line of code taking into account the delivery. It is worth mentioning that developers
context. Earlier, coders do not have such an option. save time and allocate more time to brainstorming
They were supposed to write at length what they and innovation by doing away with repetitive tasks
needed to build functional components of the (Maninger et al., 2024). Technology firms relying
software without real-time support. But, now with on AI tools for coding and code generation also
AI, it has become possible. Now, many gain competitive advantages as their teams become
programming platforms have implemented AI much smarter with AI assistance and begin to
intending to support developers in developing the produce code at a great speed which helps complete
code they need with suggestions and projects on time. One of the key aspects of an AI-
recommendations (Padmanaba et al., 2019). driven environment of coding is that it enhances
team collaboration, brings different teams from
AI Assistants: Convenience vs. Risks different platforms to one single and unified
platform where all the team members, leaders, and
According to Liang et al. (2024), there has been a decision-makers can find it easy to code in
great increase in dependence on AI assistants. A collaboration, and develop various components of
quantitative study on software development the software without compromising on quality. This
practices and the use of AI assistants such as collaborative approach not only reduces the time
GitHub Copilot and OpenAI's ChatGPT and other needed for the development of the product but also
similar tools that provide transformative insights makes the product more powerful. Research also
indicates how developers prefer to consult one of showed that AI and Human collaboration yield
such tools to code programs and how such tools better results. AI-generated code could create
also help increase efficiency, quality, and problems if it is not paired with human intelligence.
productivity. It is a well-known fact that AI tools Refining, validating, and rewriting it to fit the
are a little biased in producing code that could be purpose is crucial. Research emphasizes the
insecure too as tools are trained without much due importance of humanized systems which ultimately
attention to the sources. The sources for the data help with enhancing software productivity and
could be verified or unverified. Therefore, there is a reliability (Perry, et al., 2023).
high risk of code going wrong, which might disrupt
the work of a developer who depends entirely on Testing With AI
the AI output for the code. Apart from this, research
conducted by Pinto et al. (2024) demonstrate how AI now plays a critical role in testing. Software
concerns are prevailing in areas of plagiarism and developers find it easy to generate and execute a
copyright as the code generated by such tools could wide variety of tests, which help identify bugs. AI
match the existing pieces of code or copyrighted algorithms are employed to effectively overcome
programs. Apart from this limitations in the challenges that largely impact the overall software

© 2025, JOIREM |www.joirem.com| Page 3

 Journal Publication of International Research for Engineering and Management (JOIREM)
Volume: 10 Issue: 01 | Jan-2025

quality. Study conducted by Heusser and Larsen data is one big problem that escapes attention of the
(2023) show that with proper analysis, potential organizations. AI systems trained on non-verified
vulnerabilities are weeded out. This in turn databases might produce unsuitable results.
enhances the testing lifecycle efficiency. It is Developers using such output might be at risk of
applied to enhance exploratory testing processes. developing a system that doesn't align with the
Similarly, it is employed to analyze the data, client project goals or at least it will disrupt the
cases, and development patterns. Furthermore, AI- development process or might increase the time for
trained systems and processes reduce the time production (Hossain Faruk et al., 2022).
needed to heal the programs as they can
automatically detect the bugs and fix the systems. Implementing AI in SDLC requires a considerably
Kasowaki and Akara (2023) describe AI-driven greater initial investment. Enterprises might have to
systems as more adaptive as they can adapt to bear the losses if it doesn't deliver the expected
ongoing software development processes too, and results. According to Khaliq, Z et al. (2022)
suggest what is working and what is not. Apart technically complex would be the process of
from this, AI synced with cloud technology allows integrating AI into legacy software development
the users to scale systems up and down as required systems. It is one big issue that hinders
with extensive modifications in codebases. organizations to experiment or innovate. Hence,
companies prefer to revert to their old systems or
With the resurgence of AI in software development, methodologies of developing the software.
corporate firms prefer to employ AI Agents for Furthermore, Fischer et al. (2022) stressed the
development. Companies employing AI systems to importance of developing AI models that could
develop different components of the software notice sync perfectly well with the SDLC and other
a greater level of efficiency in all phases of SDLC. integrated systems but creating such models would
Each AI agent will be responsible for carrying out need unparalleled expertise. A lack of technical
its process of design, coding, re-coding, fixing the understanding of how AI could be implemented to
code, debugging, testing, and management. Though streamline the SDLC might stop organizations from
it is considered risky, technological advances, adopting it. Machine intelligence can't be superior
particularly in AI will make this vision more to human intelligence. Therefore, relying only on
achievable and this will further reduce the necessity the analyses generated through AI systems is never
of hiring multiple employees for the task (Donvir et recommended as such analytics might lead to unfair
al., 2024). Even if AI moves to this level, there will results or might disrupt the process of development.
be a need for human intervention as this will help
organizations ensure that the products are designed, Leveraging AI’s Potential
developed, and tested as per the standards and it is
ready for use or to make an impact (Korzeniowski AI systems are more vulnerable to attacks and they
and Goczyła 2019). could malfunction if they are fed with the wrong
data. The scalability of such systems is also one big
problem since it invites investment from time to
Transformative Role of AI in Development:
time to comply with modern development in this
Current Challenges and Possible Solutions field. Managing the systems also adds to the
increasing costs (Prather et al., 2024). Though there
Though AI plays a transformative role in SLDC are small and big challenges in implementing and
there are challenges that it grapples with. AI models integrating AI into SDLC to achieve much better
are trained on huge datasets but still, poor-quality results faster, they could be mitigated with proper
planning. Promoting AI literacy should be the first

© 2025, JOIREM |www.joirem.com| Page 4

 Journal Publication of International Research for Engineering and Management (JOIREM)
Volume: 10 Issue: 01 | Jan-2025

step as knowledgeable employees with a sound generating code snippets, deploying, testing, debugging,
understanding of AI in SDLC can ultimately and management. It is possible that shortly AI systems
outperform in development. Organizations should integrating with software development technologies will
focus on small projects first where they could use open new ways of autonomous coding. It will be
the AI to discover its potential and the inherent possible to generate completely error-free code with
descriptions in natural language. The developers will
challenges and as they reach new milestones of have more time to resolve problems and innovate while
success they can move to bigger projects. the AI systems will simplify small and big repetitive
Furthermore, problems existing in core areas such tasks. Manual coding will be reduced to a greater extent.
as communication, collaboration, and project Personalizing the software programs with AI is limited
management can be resolved with the integration of now but training verified data systems will power the AI
modern M2M (machine-to-machine) strategies. systems greatly and AI systems could dynamically adapt
Building such M2M APIs could be challenging but to software development processes and yield better
it will have the potential to facilitate results. Apart from this, by reducing errors, saving time,
communications greatly as the agent could interact and improving the collaboration among teams, it will
seamlessly. An agent developing the code will continue to be the first choice for the SDLC
share the code with the agent responsible for management.
analysis. As the code is analyzed it could be shared
with the agent responsible for debugging. AI REFERENCES
agents will work perfectly well when there is
human intervention. Though the AI agents will act
autonomously, a human agent responsible for the 1. Hernández-Ledesma, G., Ramos, E. G., y
management of the process will have a bigger Fernández, C. A. F., Aguilar-Cisneros, J. R.,
responsibility. There are bigger challenges in Rosas-Sumano, J. J., & Morales-Ignacio, L. A.
(2017). Selection of Best Software Engineering
developing such multi-agent platforms where these Practices: A Multi-Criteria Decision Making
AI-driven agents can take the task of developing the Approach. Res. Comput. Sci., 136, 47-60.
software on their own. There can be challenges in
terms of defining the roles and objectives for every 2. Farley, D. (2021). Modern Software
agent. Designing AI architectures and developing Engineering: Doing What Works to Build Better
systems with different capabilities might pose Software Faster. Netherlands: Pearson
bigger challenges for the firms but with proper Education.
planning, organizations could achieve these goals
and thus streamline their software development 3. Seemann, M. (2021). Code That Fits in Your
processes. No doubt, combining the output of AI Head: Heuristics for Software Engineering.
United Kingdom: Pearson Education.
with human intelligence should be prioritized as it
is the only way to generate flawless solutions. 4. Pantiuchina, J., Mondini, M., Khanna, D.,
Without human intervention, AI technology in Wang, X., & Abrahamsson, P. (2017). Are
programming solutions might sound meaningless or software startups applying agile practices? The
superficial. Updating the AI-driven systems should state of the practice from a large survey. In
be a top priority concern. Continual data training Agile Processes in Software Engineering and
will make the AI system more powerful and this Extreme Programming: 18th International
way it will generate the right output and software Conference, XP 2017, Cologne, Germany, May
development will be an error-free process. 22-26, 2017, Proceedings 18 (pp. 167-183).
Springer International Publishing.
3. CONCLUSIONS
5. Barenkamp, M., Rebstadt, J., & Thomas, O.
Though AI has been in use for over two decades and has (2020). Applications of AI in classical software
successfully transformed the world of technology, it still engineering. AI Perspectives, 2(1), 1.
has a long way to go. With the development of more
modern super-intelligent systems, it could be possible to 6. Amershi, S., Begel, A., Bird, C., DeLine, R.,
automate software engineering or at least core activities Gall, H., Kamar, E., ... & Zimmermann, T.
that fall within areas of design, recoding, reuse of code, (2019, May). Software engineering for machine

© 2025, JOIREM |www.joirem.com| Page 5

 Journal Publication of International Research for Engineering and Management (JOIREM)
Volume: 10 Issue: 01 | Jan-2025

learning: A case study. In 2019 IEEE/ACM 41st IEEE/ACM International Conference on

International Conference on Software Software Engineering (pp. 1-13).
Engineering: Software Engineering in Practice
(ICSE-SEIP) (pp. 291-300). IEEE. 17. Pinto, G., De Souza, C., Rocha, T., Steinmacher,
I., Souza, A., & Monteiro, E. (2024, April).
7. Marar, H. W. (2024). Advancements in software Developer Experiences with a Contextualized
engineering using AI. Computer Software and AI Coding Assistant: Usability, Expectations,
Media Applications, 6(1), 3906. and Outcomes. In Proceedings of the
IEEE/ACM 3rd International Conference on AI
8. Shneiderman, B. (2020). Bridging the gap Engineering-Software Engineering for AI (pp.
between ethics and practice: guidelines for 81-91).
reliable, safe, and trustworthy human-centered
AI systems. ACM Transactions on Interactive 18. Maninger, D., Narasimhan, K., & Mezini, M.
Intelligent Systems (TiiS), 10(4), 1-31. (2024, April). Towards Trustworthy AI
Software Development Assistance. In
9. Philipson, G. (2004). A short history of Proceedings of the 2024 ACM/IEEE 44th
software. In Management, Labour Process and International Conference on Software
Software Development (pp. 13-44). Routledge. Engineering: New Ideas and Emerging Results
(pp. 112-116).
10. Mijwel, M. M. (2015). History of Artificial
Intelligence Yapay Zekânın T arihi. Computer 19. Perry, N., Srivastava, M., Kumar, D., & Boneh,
Science,(April 2015), 3-4. D. (2023, November). Do users write more
insecure code with AI assistants?. In
11. Vaidya, J., & Asif, H. (2023). A Critical Look at Proceedings of the 2023 ACM SIGSAC
AI-Generate Software: Coding with the New AI Conference on Computer and Communications
Tools is Both Irresistible and Dangerous. Ieee Security (pp. 2785-2799).
Spectrum, 60(7), 34-39.
20. Heusser, M., & Larsen, M. (2023). Software
12. De Silva, D., & Alahakoon, D. (2022). An Testing Strategies: A testing guide for the
artificial intelligence life cycle: From 2020s. Packt Publishing Ltd.
conception to production. Patterns, 3(6).
21. Kasowaki, L., & Akara, N. (2023). Exploratory
13. Liu, B., Li, G., Zhang, H., Jin, Y., Wang, Z., & Testing Strategies for Software Quality
Shao, D. (2024). The Gap Between Trustworthy Assurance (No. 11360). EasyChair.
AI Research and Trustworthy Software
Research: A Tertiary Study. ACM Computing 22. Donvir, A., Panyam, S., Paliwal, G., & Gujar, P.
Surveys, 57(3), 1-40. (2024, October). The Role of Generative AI
Tools in Application Development: A
14. Kulkarni, R. H., & Padmanabham, P. (2017). Comprehensive Review of Current
Integration of artificial intelligence activities in Technologies and Practices. In 2024
software development processes and measuring International Conference on Engineering
the effectiveness of integration. Iet Software, Management of Communication and
11(1), 18-26. Technology (EMCTECH) (pp. 1-9). IEEE.

15. Padmanaban, P. H., & Sharma, Y. K. (2019). 23. Korzeniowski, Ł., & Goczyła, K. (2019).
Implication of Artificial Intelligence in Software Artificial intelligence for software development:
Development Life Cycle: A state of the art the present and the challenges for the future.
review. 2019 IJRRA all rights reserved. Biuletyn Wojskowej Akademii Technicznej,
68(1).
16. Liang, J. T., Yang, C., & Myers, B. A. (2024,
February). A large-scale survey on the usability 24. Hossain Faruk, M. J., Pournaghshband, H., &
of AI programming assistants: Successes and Shahriar, H. (2022, October). AI-oriented
challenges. In Proceedings of the 46th software engineering (AIOSE): challenges,
opportunities, and new directions. In

© 2025, JOIREM |www.joirem.com| Page 6

 Journal Publication of International Research for Engineering and Management (JOIREM)
Volume: 10 Issue: 01 | Jan-2025

International Conference on Software Process

Improvement (pp. 3-19). Cham: Springer
International Publishing.

25. Khaliq, Z., Farooq, S. U., & Khan, D. A. (2022).

Artificial intelligence in software testing:
Impact, problems, challenges and prospect.
arXiv preprint arXiv:2201.05371.

26. Fischer, L., Ehrlinger, L., Geist, V., Ramler, R.,

Sobiezky, F., Zellinger, W., ... & Moser, B.
(2020). Ai system engineering—key challenges
and lessons learned. Machine Learning and
Knowledge Extraction, 3(1), 56-83.

27. Prather, J., Reeves, B. N., Leinonen, J.,

MacNeil, S., Randrianasolo, A. S., Becker, B.
A., ... & Briggs, B. (2024, August). The
widening gap: The benefits and harms of
generative ai for novice programmers. In
Proceedings of the 2024 ACM Conference on
International Computing Education Research-
Volume 1 (pp. 469-486).

28. Santa Barletta, V., Cassano, F., Pagano, A., &

Piccinno, A. (2022, November). New
perspectives for cyber security in software
development: when end-user development
meets artificial intelligence. In 2022
International Conference on Innovation and
Intelligence for Informatics, Computing, and
Technologies (3ICT) (pp. 531-534). IEEE.

29. Artificial Intelligence Methods For Software

Engineering. (2021). Singapore: World
Scientific Publishing Company.

30. Lean and Agile Software Development: 6th

International Conference, LASD 2022, Virtual
Event, January 22, 2022, Proceedings. (2022).
Germany: Springer International Publishing.

31. Lee, K., Qiufan, C. (2021). AI 2041: Ten

Visions for Our Future. United States: Crown.

32. Artificial Intelligence Methods For Software

Engineering. (2021). Singapore: World
Scientific Publishing Company.

© 2025, JOIREM |www.joirem.com| Page 7