Sridhar Vembu of Zoho has revised his mind, stating that software engineers won't be replaced by AI

AI Can Boost Coding but Can’t Replace Software Engineers: Zoho Founder Sridhar Vembu

 Zoho founder and chief scientist Sridhar Vembu said artificial intelligence (AI) can significantly enhance coding productivity but cannot replace software engineers, stressing that human judgement remains essential to “orchestrate” AI systems and achieve meaningful outcomes.

Speaking at a tech townhall inside Zoho on Friday, Vembu said that as AI increasingly acts as a coding assistant, senior engineers play a crucial role in guiding models through structured prompts, intervening when they get stuck, and refining outputs using experience and contextual understanding.

In a post on X, Vembu said Zoho engineers recently reviewed C++ code generated by Anthropic’s Claude Opus 4.5 model. C++ is widely used for building high-performance software such as operating systems, databases and large enterprise applications.

Claude Opus 4.5, unveiled in November last year by Amazon- and Alphabet-backed Anthropic, is among the most powerful models in the Claude family, known for deep reasoning, memory and coding capabilities, as well as applications in areas such as financial modelling and forecasting.

According to Vembu, the review found that much of the AI-generated code was relatively simple, with only a small portion involving genuine complexity. He also noted that the model tended to be needlessly verbose.

“I suspect that the AI-generated code tends to be needlessly verbose, but I have to study it more to be sure. On the whole, I am both impressed and not super awed. I believe we can do better,” Vembu said.

He added that AI performs particularly well at writing “glue code”—the connective software that links systems, moves data and works with application programming interfaces (APIs)—tasks that are often repetitive and time-consuming for human developers. AI also shows strong pattern recall, especially for open-source code it has effectively memorised, though this raises the risk of hallucinations, he cautioned.

Vembu’s remarks mark a shift from his comments last year, when he said AI would take over most coding tasks due to the repetitive nature of programming. Reflecting on the recent review, he said he now has a clearer understanding of where AI excels and where human engineers remain indispensable.

“I now have a much clearer understanding of what these models do well,” he said.