By Bharani Subramaniam, CTO – India and the Middle East, Thoughtworks
The prevailing excitement around AI in software development rests on a single, largely unquestioned assumption: that these tools make engineers more productive by default.
But as development timelines shrink, turning what used to be a 10-day coding sprint into a single day of execution, a different reality is emerging. AI isn’t just accelerating code generation; it is radically compressing the decision-making process, creating a new kind of cognitive overload for our engineering teams.
Before AI, the actual implementation of code was the primary bottleneck in the software development lifecycle (SDLC). Developers had weeks to collaborate, test with users, understand the data, and weigh the trade-offs of their design, before or as they were coding.
Today, AI has obliterated that buffer. The cognitive load hasn’t increased because they have to remember more lines of code; it has increased because they have to make high-stakes decisions at unprecedented speeds. When you are moving this fast, a slight miscalculation at the ‘steering wheel’ drastically alters your result.
With AI, creating a working proof-of-concept in a couple of hours is easy. But this has led to a dangerous belief among business leaders: assuming that because a demo is fast, building enterprise-grade software is a similarly quick process.
Rapid AI prototypes create a false sense of accomplishment by ignoring the ‘hidden’ requirements of a real product; things like graceful failure, network stability, and meeting accessibility standards. Taking an app to production remains incredibly complex, and the assumption that AI can remove all that complexity, is unrealistic.
While AI makes building economical and lightning-fast, it doesn’t provide a sense of direction. Iterating without intent just means building the wrong thing at a higher frequency. As an industry, we must realize that AI has shifted the ultimate bottleneck from how we build software back to what we are building.
Tech leaders must spend equal, if not more, time iterating on product design and user experience as they do on code generation.
Ultimately, AI should not be restricted to the coding phase but we must deploy these tools across QA, product design, and UX to help those teams keep pace. While AI can help us build things faster than ever before, the overarching question of the AI era remains: are we building the right things?