Hyderabad: Generative AI is accelerating software development, but its efficiency comes with caveats, industry experts said at a roundtable discussion hosted by IIIT Hyderabad. While AI-powered tools can speed up coding and automate repetitive tasks, their limitations — ranging from unreliable outputs to security concerns — mean they cannot yet function without human oversight.The discussion brought together researchers, technologists and industry leaders from banking, healthcare, SaaS, startups and open-source communities. These experts examined how AI is reshaping the development process.A White Paper emerging from the session outlined how AI was helping developers cut down search times and complete tasks faster.However, panellists noted that while AI works well for straightforward coding, it struggles with legacy systems, complex logic and business-specific applications. “Responses from AI tools are incorrect nearly 50-60 per cent of the time when it comes to business logic,” one participant observed.Another growing concern was ‘technical debt’, where AI-generated code is functional but not optimised. “The challenge isn’t just about getting code out quickly, but ensuring it doesn’t create long-term inefficiencies,” an industry expert noted.The panel remained divided on whether technical debt could be eliminated or if managing it was simply a necessary trade-off.Despite the enthusiasm around AI, panellists pointed out that actual adoption among developers remains low — around 20 per cent — even in companies investing heavily in the technology.Many developers are sceptical, believing they can write better code themselves. Others worried about job security, even though experts repeatedly stressed that AI was meant to assist rather than replace human programmers.Data privacy was another major roadblock, with companies hesitant to share proprietary information with AI models due to security and regulatory concerns.Organisations that have successfully integrated AI into their workflows have done so cautiously, treating it as a junior developer rather than an autonomous problem-solver. Some banking firms, according to participants, have achieved up to 85 per cent accuracy in AI-generated SQL queries when human reviewers were involved. “If you treat AI like an intern that needs supervision, it works. But leaving it unchecked leads to inconsistencies,” a tech expert explained.Training was also flagged as a critical factor. Developers not familiar with AI tools often struggle to write precise prompts, leading to unreliable outputs. Some organisations have conducted hands-on workshops and pilot projects to refine AI adoption.Open-source community Swecha, which tested AI-assisted coding against traditional methods, found that AI-supported teams completed tasks nearly twice as fast. However, concerns remain that over-reliance on AI could weaken core programming skills over time.While AI is transforming development practices, panellists agreed that it is unlikely to reduce the overall size of engineering teams. Instead, it is changing the nature of software development, shifting focus from routine coding to higher-level problem-solving.Some experts suggested that new roles — such as AI-driven product managers — might emerge to oversee AI-generated code and ensure it meets quality standards. Companies are also working to establish governance frameworks to regulate AI use, particularly in areas of security, intellectual property and accountability.“Trust but verify” was a recurring theme, the view that AI-generated code must be reviewed before integration. As businesses adapt to AI-driven software development, the key takeaway from the discussion was that AI is a tool, not a shortcut.
Source link
