How AI is transforming the development lifecycle [Q&A]

A shortage of skilled developers is leading to many companies failing to meet their product roadmap targets, according to research from Full Scale.

As a result the software development lifecycle (SDLC) is undergoing a fundamental shift, that includes using artificial intelligence to cut out repetitive tasks and allow developers to concentrate on their core role.

We spoke to Neeraj Abhyankar, VP, data and AI at R Systems, to find out how developers can better deploy AI in the SDLC and the advantages it can offer businesses.

BN: How can AI help in ensuring the completeness of initial software requirements?

NA: AI helps uncover what’s often missed during early scoping. We’ve used AI tools to analyze past project data, stakeholder inputs, and documentation to surface hidden or implied requirements. These tools can auto-generate user stories, flag functional gaps, and build traceability matrices that link requirements to test cases and design elements, which reduces rework and improves alignment with our customers’ business goals.

However, AI still can’t fully grasp nuanced business logic or shifting priorities, which is why it’s critical that organizations continue to pair AI’s speed and breadth with human judgment to ensure requirements are both complete and contextually accurate.

BN: What’s the most significant boost to developer productivity provided by AI code assistants and where do they still fall short?

NA: The biggest gain is how AI assistants reduce repetitive work — boilerplate code, refactoring, and documentation. At R Systems, for example, we’ve used agentic tools to stabilize builds, optimize queries, and clean up legacy code. This frees developers to focus on architecture and problem-solving. It’s especially helpful for onboarding new engineers or accelerating prototypes. However, AI still struggles with complex logic, maintaining context across large codebases, and it can introduce subtle bugs if suggestions are accepted blindly. That’s why it’s important to treat AI as a collaborator, not a replacement. Human-in-the-loop validation is key, and it’s critical to continuously benchmark outputs to ensure quality and reliability.

BN: How is AI being integrated into security testing and identifying zero-day vulnerabilities more efficiently?

NA: AI is changing how businesses approach security testing. Using agentic frameworks that simulate attack scenarios and flag vulnerabilities before code goes live is a great place to start. AI helps with static and dynamic analysis, and it’s particularly good at spotting risky patterns and anomalies. For zero-day threats, unsupervised models can detect behavior that deviates from the norm, which gives us early warning signals.The use of custom benchmarks to test for prompt injection and data leakage in AI-infused systems is also something to consider. While AI speeds up detection and coverage, it’s not a silver bullet. Manual penetration testing still plays a critical role, and embedding AI into the CI/CD pipeline can help ensure continuous validation.

BN: What about the effect on long-term software maintenance, such as automatically patching security flaws or updating old dependencies?

NA: AI is helping shift long-term system maintenance from reactive to proactive. Modern tools can monitor codebases for outdated libraries and known vulnerabilities, and in some cases, suggest or apply patches automatically. We have, for example, observed AI copilots supporting tasks like dependency resolution, regression testing, and performance tuning, which all reduce manual effort and improve system reliability.

AI can also highlight architectural bottlenecks and propose refactoring strategies, contributing to reduced technical debt. However, automation must be governed carefully. Unchecked updates can introduce instability. Organizations can mitigate this risk using semantic versioning, automated test coverage, and human oversight to validate changes. The goal should be to build resilient systems that evolve safely and sustainably.

BN: How must the role and required skillset of a modern software engineer evolve to make effective use of these tools?

NA: Software engineers today need to be more than just coders — they need to be orchestrators of AI-driven workflows. It’s important that engineers learn prompt design, agentic system integration, and model evaluation. Additionally, understanding how to guide AI tools, validate their outputs, and embed them into scalable architectures is essential. Engineers also need to be comfortable with APIs, semantic modeling, and orchestration frameworks.

Outside of technical skills, critical thinking, ethical reasoning, and collaboration are just as important. There’s been an ongoing shift toward modular design and human-in-the-loop systems, which means engineers must architect for both automation and oversight.

Image credit: BiancoBlue/Dreamstime.com