The outlook for software development in 2025 – Software
The software development industry is experiencing an unprecedented pace of change, with rapid advancements in programming languages, architectures, frameworks, methodologies, and more.
At the heart of this transformation is the growing intersection of artificial intelligence (AI) and software development. Natural language programming, in particular, is helping bridge the gap between developers and non-technical users, fostering increased collaboration in the creation of software.
AI is no longer just a tool – it’s driving a revolution in the way software is built. Developers are leveraging AI to automate repetitive tasks, generate code, detect and fix bugs, and streamline the DevOps process, making continuous integration and delivery (CI/CD) more efficient than ever.
The impact of AI and growth of low-code development (increasingly AI-driven) also throws the question – how will AI change the role of the software engineer or what must they do to remain relevant?
What lies ahead in the outlook for software development in 2025? To better read into the future of the industry, we speak with industry experts across the Asia Pacific, many who are driving change, to get their perspectives on the trends, challenges, and innovations that will shape the next era of software engineering. Our respondents include:
Thomas Chia, Chief Technology Officer, Chocolate Finance ;
Diego Lo Giudice, Vice President, Principal Analyst and John Bratincevic, Principal Analyst, Forrester;
Shanker Selvadurai, Chief Technology Officer, APAC, IBM;
Jornt Moerland, SVP, APAC, Mendix;
Jan Wuppermann, Senior Vice President, Strategy & Business Operations, APAC, NTT DATA;
Leonard Tan, Regional Director for Singapore, Malaysia, Brunei and the Greater China region at OutSystems;
Sarah Taraporewalla, Chief Technology Officer for APAC, Thoughtworks
iTNews Asia: How transformative has been the role of AI? How has AI reshaped how developers code and maintain software? What specific areas have AI improved developer productivity or changed the way software is built?
Selvadurai (IBM): Advanced AI-driven solutions have revolutionised software development by automating code generation, debugging, and streamlining CI/CD pipelines. Intelligent code assistants now offer context-aware recommendations and leverage natural language processing, enabling non-technical stakeholders to engage in the development process.
This enhanced collaboration accelerates innovation while improving code quality and productivity. Additionally, these tools extend their benefits to legacy systems, modernising critical applications without compromising security or performance.
Wuppermann (NTT): AI has transformed software development, making it faster, efficient, and intelligent. AI-powered code assistants, such as generative AI models, have streamlined coding by suggesting context-aware code snippets, automating error detection, and even generating entire code structures. This has significantly reduced development time and improved code quality. For instance, NTT DATA’s Generative AI-fuelled code generation has accelerated application development, leading to productivity gains of up to 25 percent.
Additionally, AI-driven testing and debugging tools enable rapid identification of vulnerabilities and inconsistencies, accelerating software release cycles.
Lo Giudice (Forrester): AI is reshaping the industry – the integration of AI into software development can increase development teams productivity by 10 to 20 percent.
These AI-infused tools, which Forrester calls TuringBots, transition from the traditional “AppDev” to an “AppGen” approach, automating asset generation across the Software Development Life Cycle (SDLC) and enhancing efficiency. They enhance bug detection, code analysis, and understanding, improving software quality. They are also disrupting the legacy modernisation market offering real opportunities that did not exist, in migrating more quickly old legacy applications to newer architectures.
Tools like GitHub Copilot facilitate automated code generation, allowing developers to focus on complex tasks by reducing time spent on routine code. Additionally, predictive analytics help in project management by forecasting project timelines and identifying potential bottlenecks, while automated testing ensures higher quality standards by covering more scenarios.
Tan (OutSystems): Seasoned developers increasingly rely on coding assistants to generate real-time code suggestions and streamline maintenance tasks. These tools cater to a more traditional development environment where precision and efficiency are paramount.
At the same time, integrating AI with low-code platforms is changing the way software is built. Rather than simply suggesting code, AI now helps to manage tasks by understanding the development context. It simplifies complex tasks, foresees potential issues, and adjusts support based on how developers work while ensuring that human oversight remains a key part of the process.
– Leonard Tan, Regional Director for Singapore, Malaysia, Brunei and the Greater China region at OutSystems
Moerland (Mendix): AI is reshaping the enterprise software development space, focusing on driving productivity, critical thinking, and streamlining the overall development cycle.
One major shift is the role of human developers. While human oversight will always be critical, developers will transition towards leads who orchestrate “synthetic developers” and review generated code, ensuring it meets their company’s security and internal development policies.
The potential of AI is well placed to also augment low-code development capabilities, allowing teams to accomplish more with the same headcount.
Taraporewalla (Thoughtworks): While it’s still relatively early in the journey, GenAI is already redefining how developers write and maintain software by taking on routine tasks and accelerating feedback loops. We are seeing a fundamental shift: coding assistants like GitHub Copilot and emerging agentic IDEs help developers focus on designing robust architectures, ensuring security, and delivering real business outcomes.
Although Generative AI (GenAI) for coding is the most widely cited use case, actual coding accounts for only about 30 percent of a developer’s time. Far more compelling is how GenAI can be applied across the entire software lifecycle – from backlog refinement to threat modeling, where it not only saves time but also sparks innovation and creativity.
GenAI adoption can be accelerated by working alongside any AI coding assistant and adding pluggable knowledge packs to streamline development tasks, identify process gaps, and drive innovation, while also addressing intellectual property concerns.
Chia (Chocolate): In software development, it’s clear that AI, especially GenAI, is great at generating boilerplate code from high-level business prompts. Since software development often involves creating models for both static structures and operations, GenAI can suggest designs for these models and turn them into actual code.
However, maintenance is more complicated. We’re in a phase where code is created and updated by both humans and AI, and there’s still a mix of both in the process. We’re not yet at a balanced point, and it’s unlikely that the future will have humans mostly maintaining AI-generated code. We will need to embrace a new paradigm of “by the machines, for the machines”.
– Thomas Chia, Chief Technology Officer, Chocolate Finance
iTNews Asia: Do you see challenges in using AI (and GenAI), especially in scaling to large-scale enterprise applications? With AI systems expected to access large amounts of sensitive data, how can we address concerns over information security and data privacy?
Taraporewalla (Thoughtworks): There are five critical gaps that must be addressed to successfully scale AI for large-scale enterprise applications:
- Data Gaps: Data should be readily accessible, properly tagged, and enriched with the necessary metadata for each use case.
- Infrastructure Gaps: Organisations must ensure infrastructure can scale with user demand while effectively managing token-related costs.
- Evaluation Gaps: Robust methods are needed to measure performance and reliability of AI models-particularly large language models (LLMs).
- AI Skill Gaps: Teams must be prepared to embrace AI, overcome resistance to change, and understand how AI augments rather than replaces their roles.
- Adoption Gaps: Strong change management strategies are essential to guide AI adoption and generate lasting organisational buy-in.
Addressing ethical challenges includes adopting stringent data governance – with techniques like anonymisation, encryption, and differential privacy to safeguard sensitive information. Ensuring accountability involves transparent, explainable AI systems supported by clear audit trails, while diverse datasets and fairness-aware algorithms help mitigate bias.
Ethical principles can be woven into platform development by leveraging open-source AI frameworks, promoting responsible innovation through fostering transparency, offering shared resources for bias detection and security, and catalysing cross-industry collaboration.
Lo Giudice (Forrester):
Companies face challenges in scaling AI and GenAI for large-scale enterprise applications , primarily due to the need for substantial computational resources and concerns over information security and data privacy.
– Diego Lo Giudice, Vice President, Principal Analyst, Forrester
To address these challenges, many TuringBot vendors of this technology offer contractual commitment and technical solutions to avoid using customer data to train TuringBots underlying models. Customer data is used only at inference time to generate code or other SDLC assets.
Players, like IBM, MS, GitHub also offer indemnification. For some companies, cybersecurity alone may not be enough. They might need to use strict access controls, data encryption, and anonymisation techniques to protect sensitive information, especially when TuringBot vendors don’t guarantee the necessary privacy and data protection.
Chia (Chocolate): One major challenge in scaling GenAI is cost. With the current token-based pricing model, we must be mindful of rising costs, especially when external users send tokens during conversations.
Another issue is that GenAI is non-deterministic, meaning the model can produce different results from the same input and in large-scale scenarios, this can lead to errors piling up.
Security issues can be addressed with standard measures like access control and data hygiene, treating the GenAI system like any other entity in your organisation to prevent unnecessary data access.
It is also essential that the people using the tools understand the risks. This starts with knowing how prompts are handled- whether they’re stored on the model provider’s server and used to train the model, or if they’re isolated to protect user privacy.
Moerland (Mendix): Since AI depends on quality data, effective data management is critical and complex. Knowledge bases and retrieval-augmented generation (RAG) techniques improve accuracy, but IT teams must also ensure that AI systems adhere to strict security, regulatory, and data-sharing policies to prevent misuse or breaches.
Integrating AI into existing IT infrastructure, particularly legacy systems, will be a considerable undertaking. Disparate data and silos increase costs and risks while reducing agility.
To navigate these complexities, enterprises must continuously meet industry certifications, standards, and compliance requirements, both globally and locally.
Tan (OutSystems): Integrating AI into existing systems can create additional complexities, from updating legacy infrastructure to managing implementation backlogs and addressing a shortage of skilled talent. Yet, the market’s speed demands that enterprises find agile solutions to stay competitive.
As a result, businesses need more than a standalone micro-service, and low-code platforms enable the rapid creation of fully integrated, secure solutions that adapt as quickly as the market evolves.
By embedding enterprise-grade security and compliance into every step of the development process, organisations can protect sensitive data while also streamlining code governance and system updates.
Wuppermann (NTT): Ensuring model reliability, interpretability, and robustness at scale requires significant computational resources and governance frameworks. AI systems must comply with stringent regulatory standards to handle sensitive enterprise data securely.
Organisations should implement robust data governance strategies, including encryption, access controls, and federated learning approaches that enable AI models to be trained without directly accessing sensitive data. AI ethics frameworks and transparency in AI decision-making also play a crucial role in building trust and mitigating risks.
– Jan Wuppermann, Senior Vice President, Strategy & Business Operations, APAC, NTT DATA
Selvadurai (IBM): Advanced solutions must incorporate robust encryption, identity management, and real-time threat detection to protect sensitive data. Adhering to privacy-by-design principles, techniques such as data anonymisation and secure computation can ensure compliance with global regulations.
Successful implementations in complex legacy and hybrid environments demonstrate that with rigorous governance, these AI systems can be scaled effectively to meet enterprise needs.
iTNews Asia: How do you see “no-code” and “low-code” approaches impacting the future of software development in 2025? Are they ready to be deployed across large scale mission critical applications? What advantages can such programming provide organisations with, and how can we ensure proper governance, security and compliance?
Chia (Chocolate): LCNC platforms offer great potential, but they should be considered within the broader business context. While they can boost productivity, they follow a different approach to software development, which comes with certain limitations. These platforms are designed to improve efficiency work inherently by reducing the number of possibilities, so they may not be as flexible or comprehensive as traditional software development teams.
Using no-code/low-code for large-scale apps requires careful thought. While they enable agility, their limitations must be weighed against business needs. Additionally, LCNC tools still rely on someone else’s code, so there are risks related to code quality, security, and maintainability. To reduce these risks, it’s crucial to separate low-code applications from critical infrastructure to protect an organisation’s overall security.
Taraporewalla (Thoughtworks): LCNC platforms can speed up time-to-market, allow “citizen developers” to quickly prototype solutions, and reduce reliance on scarce development skills. However, there are potential risks to watch out for, such as vendor lock-in, limited customisation, governance issues like “shadow IT” and the possibility of weakening good engineering practices if overused for core systems.
In short, no-code/low-code is a useful tool for specific tasks, like testing ideas or digitising simple workflows, but organisations should understand their limits, especially around customisation, data security, and long-term maintenance.
– Sarah Taraporewalla, Chief Technology Officer for APAC, Thoughtworks
Wuppermann (NTT): LCNC platforms are making software development more accessible, allowing business users to create applications with little technical expertise. These platforms cut development time, costs, support rapid prototyping, and integrate easily with enterprise systems.
They are now evolving to support more complex, large-scale applications, but they are not a one-size-fits-all solution for mission-critical systems. For instance, at NTT DATA, we are building our own Automation & GenAI platform using LCNC.
To maximise their potential, organisations must implement best practices such as role-based access controls, continuous security monitoring, and compliance audits. Hybrid approaches, where LCNC is combined with traditional coding for customisation and performance optimisation, can further enhance their viability for enterprise-wide applications.
Moerland (Mendix): LCNC will accelerate digital transformation, enabling enterprises to build high-quality apps cost-effectively at speed and scale. We’ve seen complex, mission-critical software being developed as much as six times faster with low-code versus traditional development methods, improving agility and time-to-market while reducing resource investment by up to 70 percent.
While both low-code and no-code lower technical barriers, they serve different needs. No-code empowers business users to build simple apps with minimal IT involvement, while low-code supports complex, enterprise-grade applications by integrating with IT infrastructure ensuring governance, security, and compliance.
Successful LCNC adoption requires a holistic approach that addresses people, processes, and tools. Enterprises must outline clear usage guidelines, enforce security protocols, and ensure regulatory compliance for effective data management and security.
Selvadurai (IBM): Low-code/no-code (LCNC) platforms are proving pivotal in democratising application development. When integrated with intelligent automation tools, these platforms empower both technical experts and citizen developers to rapidly build enterprise-grade applications.
LCNC platforms reduce development time and costs while ensuring robust governance, security, and compliance. Mission-critical applications can be built using these frameworks, combining agile innovation with the strict standards needed for large-scale deployments.
Bratincevic (Forrester): Low-code development is faster, easier, and cheaper than high-code, even with the influence of AI. It is commonly used for large scale, mission critical applications such as core banking and personal banking experiences, insurance policy management, ERP, manufacturing execution systems, large field operations applications, etc.
We expect software development to become more low-code in the next few years as “Application Generation Platforms” (AppGen) evolve. These platforms use AI throughout the software development process, streamlining traditional steps and making developers more versatile in their roles.
– John Bratincevic, Principal Analyst, Forrester
To ensure proper governance, security, and compliance, a dedicated team like a centre of excellence must lead the low-code strategy, providing training, mentoring, and automation throughout the application life cycle.
Tan (OutSystems): Low-code, unlike no-code, offers more flexibility and customisation for organisations to build applications. It has helped businesses tackle challenges like modernising legacy systems, closing the talent gap, and improving data security. With the addition of AI, these platforms are speeding up development cycles and making it easier to adapt to market demands.
Although AI-powered low-code tools are becoming more capable for mission-critical applications, governance, security, and compliance are still crucial. Business leaders are turning to enterprise-grade low-code platforms, which often include built-in security checks, disaster recovery, and compliance features to ensure data integrity.
iTNews Asia: Do you see the role of Citizen Developers (non-technical users building apps using low-code/no-code tools) growing in APAC, much like the US? Are there ways AI can streamline the collaboration between technical and non-technical teams when developing enterprise-level software solutions?
Wuppermann (NTT): The rise of citizen developers in APAC mirrors global trends, as businesses see the value in empowering non-technical users with LCNC tools. Gartner predicts that by 2026, 80 percent of low-code app users will be from outside IT departments, showing a shift towards business-led development.
AI can boost this collaboration by offering real-time guidance, automating workflows, and converting business logic into optimised code. AI-driven platforms help bridge the gap between technical and non-technical teams with natural language interfaces, automated quality checks, and security recommendations, ensuring that citizen developers create apps that meet enterprise standards while promoting innovation across the organisation.
Chia (Chocolate): The rise of Citizen Developers is growing worldwide, with organisations in digital-first regions seeing increasing demand for user-driven app development. This is especially true in sectors like fintech, where speed to market is crucial.
LCNC platforms allow non-technical users to create solutions that fit their business needs without waiting for lengthy development cycles. This trend has been boosted by the evolution of GenAI, which not only helps generate LCNC products but can also be integrated into no-code/low-code workflows.
While I believe that effective collaboration still depends on human interaction, I see more non-technical stakeholders taking a hands-on approach, which helps demystify the complexities of building digital systems beyond the engineering team.
Moerland (Mendix):The growing demand for digital applications outpaces the availability of skilled developers, making citizen development a key trend in APAC.
However, success of citizen development also depends on delivering the right software, which requires collaboration between technical and non-technical teams. Low-code platforms support this, and AI’s generative capabilities can enhance development further. For example, development teams can use AI tools to gain real-time, context-based recommendations during development, leading to more innovative and impactful software.
Tan (OutSystems): Many citizen-developed applications are siloed and simple, often lacking integration with the broader IT ecosystem. This can lead to inefficiencies, as ideas may need to be rewritten on other platforms, defeating the “concept to code” goal. This is where AI comes in.
When combined with low-code tools, AI helps technical teams speed up development while allowing non-technical teams to contribute to planning and ideation. It ensures that applications integrate smoothly into the larger IT system.
For instance, features like coding assistants can quickly turn business requirements into templates and app designs, to create a strong initial draft in just a matter of minutes. However, this initial draft is just the beginning and will be further refined.
Taraporewalla (Thoughtworks): GenAI-powered tools empower citizen developers, but they also bring risks like Shadow AI, where decentralised solutions create governance and integration challenges. While these tools speed up the ideation phase, there are still concerns about their maintainability and scalability as they grow.
To fully unlock their potential, tech teams should focus on automated guardrails, compliance-as-code, and building strong platforms that balance agility with enterprise standards. However, fostering responsible AI goes beyond technology alone – it requires strong processes, a supportive culture, and close collaboration between technical and business stakeholders.
Bratincevic (Forrester): We see there is an increase in need for citizen developers who understand business domains and can work effectively with large language models (LLMs) through prompt engineering and lightweight retrieval augmented generation (RAG) to deliver AI-infused apps and agents at scale.
AI-driven growth in citizen developers may boost overall collaboration between business and tech in aggregate for tasks that are tightly governed by IT.
Selvadurai (IBM): The rise of citizen developers is crucial to digital transformation, especially in emerging markets. With the help of easy-to-use automation tools, non-technical users are playing a larger role in application development, connecting business needs with technical execution.
This collaboration speeds up innovation and reduces IT backlogs. These tools improve communication and integration across teams, resulting in agile, high-quality enterprise software solutions.
iTNews Asia: With AI and LCNC making application development more easily accessible, how should software developers navigate the changing landscape? How should they enhance their skills and what new roles can they play to remain relevant in 2025?
Tan (OutSystems): Instead of focusing on manual coding tasks, developers will step into more strategic roles, designing AI frameworks and prompts to accelerate app generation with greater precision and agility. This shift will also see developers collaborating more closely with decision-makers to explore how technology can be utilised to better drive business impact.
To remain competitive, developers must upskill in AI, automation, and system design. Beyond technical skills, developers must also sharpen their communication and problem-solving abilities to effectively work with non-technical teams across various departments.
Bratincevic (Forrester): Professional developers must accept that they won’t be building all apps in the future. Most development will be done by business domain experts using low-code and AppGen platforms.
True technologists should focus on three areas – platform engineering (e.g., adding new services), mentoring and collaborating with citizen developers on technical tasks, and gaining skills in complementary areas of software development (e.g., product management, design, analysis) as “multidisciplinary digital problem solvers” become the norm for delivering large software solutions.
Selvadurai (IBM):
Developers should see AI and automation as growth opportunities. Continuous upskilling in areas like AI integration, cloud architectures, secure coding, and low-code/no-code platforms is crucial.
– Shanker Selvadurai, Chief Technology Officer, APAC, IBM
As the industry evolves, new roles such as AI/ML specialists, low-code architects, and digital transformation consultants are gaining prominence. Embracing these changes and expanding skill sets will help developers lead digital transformation and drive business success.
Chia (Chocolate): A good developer understands that their value lies in turning vague business needs into functional digital systems, not just typing code. GenAI and LC/NC still require the critical thinking and analysis that skilled developers bring.
Developers should adopt a “centaur” mindset, using these tools to improve the development process rather than fearing replacement. The rise of roles like “prompt engineer” highlights the ongoing need for technical expertise even in GenAI.
Taraporewalla (Thoughtworks): Developers must shift focus to architecture, integration, security, and scalable software quality. As AI assistants handle more coding tasks, building deep expertise in areas like cloud infrastructure, DevOps, and machine learning operations (MLOps) becomes essential for driving and maintaining complex solutions. They can also evolve into “AI facilitators” or “prompt engineers,” shaping how AI tools are trained and orchestrated to meet business needs while preserving ethical and responsible use.
Domain expertise will remain a competitive advantage, as developers who understand specific industries can better translate technical possibilities into real-world value.
Moerland (Mendix):
LCNC and AI are not replacing developers – they are amplifying their capabilities. In 2025, the most successful developers will be orchestrating intelligent systems that solve business problems while navigating the technical, ethical, and organisational challenges of AI-augmented development.
– Jornt Moerland, SVP, APAC, Mendix
To remain future-ready, professionals need to embrace three shifts – develop expertise in multimodal approaches that move beyond traditional tabular data thinking; understand distributed intelligence as computing power spreads to edge devices; and master the governance of AI tools.
Wuppermann (NTT): Developers should focus on areas like AI model training, advanced system architecture, cybersecurity, and DevOps automation. Expertise in integrating LCNC platforms with enterprise systems and ensuring governance will be in high demand.
Basic coding and testing tasks will become less important. Instead, learning how to prompt AI effectively is becoming more relevant. In other words, developers should become leaders of AI agents, guiding them to success.
link
