Who's at the Helm? Sulakshana Singh on Gen AI's Role as Copilot

Tab may be the coder’s new best friend, but generative AI isn't just a handy autocomplete—it's changing how software is designed, built, and delivered, top to bottom.

Sulakshana Singh, an accomplished software engineer with over a decade of full-stack development experience, has led major projects at Equifax Workforce Solutions, IBM, and Société Générale. A recipient of the 2024 IEEE St. Louis Section Award for Women in Engineering and technical peer reviewer for IEEE IoT, Singh is known for tackling the often tedious software development lifecycle (SDLC) with cutting-edge tools. We sat down with her to get an inside look at how generative AI is transforming every stage of software development and the opportunities—and challenges—it brings.

Gen AI coding tools are everywhere now, fueling the "copilot" phenomenon. What impact are they having on the SDLC?

The impressive thing is that gen AI is providing different benefits at each stage of the development cycle. Machine learning was already a staple in testing and maintenance, but gen AI is now able to add value in earlier phases like requirements gathering and design. If I had to generalize, they address bottlenecks. These tools are excellent for automating repetitive tasks and allowing engineers to channel their expertise into solving more strategic, user-focused and business-driven problems. They accelerate a lot of due diligence which ultimately helps produce higher-quality software in less time.

Let’s start from the ground up. How does gen AI improve requirements gathering?

Requirements gathering is very often one of the most important, time-intensive phases because it sets the tone for your entire project. Nowadays, tools like Jira AI can break down project prompts into subtasks so engineers don't need to perform discovery from scratch. For instance, inputting "build an online bookstore" might generate subtasks such as creating a search API or designing an ISBN-driven user interface.

NLP-powered models like ChatGPT can turn scattered ideas into actionable deliverables or user stories, while IBM Watson is able to analyze business docs and extract detailed requirements. For large, complex projects with extensive historical data, like those I’ve worked on in finance and social services, these tools can save us considerable time. Engineers are freed up to collaborate more effectively with stakeholders instead of spending hours combing through documentation.

How does AI streamline the design and architecture phase?

The design phase is all about striking a balance between business and technical requirements, and AI is a huge help in simplifying that process. Tools like Lucidchart can generate UML diagrams from simple, non-technical descriptions, and will organize system features in a fraction of the time it would take to whiteboard them. Even prototyping tools like Figma can now suggest UI designs based on provided business requirements, expediting iterations for devs collaborating with non-technical teams.

And because these tools retain a holistic view of the project, they can also optimize for cost and performance trade-offs. Our product team collaborators really appreciate what these tools bring to the table.

Coding is where AI got its start in development. How has it evolved?

Most developers are familiar with code assistants that suggest methods or subsequent functions based on your inputs. Now, they’ve advanced to generating entire microservices from API specs. This lets developers focus on implementing custom business logic instead of getting bogged down in integration tasks.

There's a massive extensibility network as well—these tools now support multiple languages and frameworks, helping teams build modular, reusable code faster. It all leads to better code quality and fewer cycles, and devs are always looking for more efficient workflows.

How does AI address testing and maintenance challenges?

It dramatically reduces the time spent on testing and release cycles, or removes the need for manual QA entirely. At the basic level, defect analysis tools can use NLP to analyze code and recommend fixes. More advanced tools, like Testim, automate GUI testing, while mutation testing powered by AI can uncover edge cases that traditional methods might not catch.

In production, it's been great for risk management. Tools like Dynatrace predict potential system failures by looking at historical data, enabling preemptive fixes and smoother follow-up deployments. Most anything with prescribed standards or guidelines can be automated.

If gen AI is so transformative, what’s slowing down its adoption?

It comes down to a mix of concerns about bias, privacy, and transparency.
From a technical standpoint, AI models are trained on historical datasets, which can introduce biases that clash with newer best practices or proprietary code. And privacy and intellectual property are an ongoing concern, especially when dealing with sensitive materials. The provenance of training data raises questions about who, if anybody, owns what's produced. Addressing these challenges will take time, and until they're resolved, many organizations are going to be cautious.

Transparency is more of a broader issue with AI. which makes accountability tricky and has fueled demand for XAI (explainable AI). But, most importantly, the lack of open-source support is a sticking point for developers, who value clarity and open-source collaboration very highly.

What excites you most about the future of generative AI in software development?

I’m thrilled to see generative AI becoming a standard feature in IDEs, making them accessible from the start of the coding process.

Beyond speeding up delivery, I'm eager to see the evolution of CI/CD pipelines. Imagine test cases updating in real-time alongside code changes, or fully autonomous maintenance systems capable of self-healing.

These innovations could deliver software that not only better meets user needs but also stays more stable over time. It’s a fascinating time to be an engineer. The tools we’re building and refining today could set a new standard for the entire industry.