AI is rewiring the software industry by collapsing traditional workflows

Autonomous agents are shortening project timelines and changing how people work in the sector, reports Avik Das

For decades, software development was governed by rigid, linear frameworks. From the traditional Waterfall model’s strict sequence of specification docs and static wireframes to iterative feedback loops, the core engineering process has always relied on step-by-step human execution. 

Now, that pipeline is being disrupted. By writing millions of lines of code and compressing design, development and testing into a single fluid operation, autonomous artificial intelligence (AI) agents are not just accelerating production — they are permanently dismantling the technology industry’s foundational software development lifecycle. 

Software development in the Waterfall model resembles water cascading down stairs. Once an engineer completes all tasks in a phase, she cannot return to that phase unless the project’s requirements change significantly. While the Waterfall model was followed by Agile in the 2000s and DevOps later, the concept remained the same, albeit with iterative development, user feedback and refinement. 

The model worked best in software projects that had clear requirements and specifications at the start of the development. The process is deterministic, offering predictability when it comes to meeting deadlines and making it the popular choice for traditional software projects. 

But generative artificial intelligence (GenAI) is changing that. While steps such as planning, analysis, design, coding, testing, deployment and maintenance will continue, GenAI can be applied across the software development lifecycle (SDLC) for quicker turnaround. A single cycle of coding could take weeks or months, but GenAI can compress the work to days. 

“In the earlier days, we wrote all the specifications, then went to the next step and then all the way down to production. And then we created some version of Agile, with each step kind of flowing into the next and it’s this continuous loop,” Kyle Daigle, chief operating officer of GitHub, Microsoft’s open-source code-hosting platform, told Business Standard in April. 

“I think when we are building software with agents, the steps are no longer clearly defined. You don’t finish code and then test it. Instead, there’s an agent testing it in the background as you’re writing it. It’s really about how do you define the steps when the steps can now happen concurrently instead of consecutively. And so we talk about build, deploy and operate as this new model because it tends to be where our focus is,” he said. 

Software developers no longer finish one task and hand it over to the next person. “Instead they’re able to move more quickly and get more done across the entirety of that SDLC,” said Daigle, who is also Microsoft’s chief marketing officer, developer.

This shift from a standard Waterfall or Agile model to conversational development is happening through vibe coding. Vibe allows people with no knowledge of computer languages to code using plain English commands. Startups like Cursor and Windsurf — the latter was acquired by Google in a $2.4 billion deal in July 2025 — have quickly become the darlings of investors seeking fresh bets in AI.   

In the pre-AI era, the wall between software development and operations was already collapsing with DevOps, says Praveen Bhadada, chief executive officer (CEO) and managing director of management consulting firm Neovay Global. 

“Steps that used to be handed off in sequence (build , test, deploy, monitor), started happening in automated loops. DevOps still assumed a human was writing the code, defining requirements, making architectural decisions and debugging logic. The sequence of cognitive work remained largely intact. Agentic AI is compressing the cognitive sequence itself as these steps can now happen near-simultaneously, initiated by one person or a team of fewer people.” 

Experts say that as SDLC stages overlap, engineers might manage five or six phases simultaneously. While traditional integrated development environments assume a sequential order, AI agents and copilots allow development to happen in any order. Ultimately, the human in the loop remains in control, determining what the agent does. 

At the same time, the merger of various lifecycle roles creates risks for human capital, potentially leading to job losses. It reshapes job descriptions, creating a demand for entirely different skill sets. Executives note that small teams of just three to four members are now capable of executing projects end-to-end, fundamentally changing both the development model and the delivery cycle. 

“Even when Waterfall moved into Agile, you would have always had someone create a backlog of product ideas, someone who will create a backlog of features that we want to develop, a development team picking that up, testing team following and then doing the whole end-to-end maintenance. That now changes because the teams are merging themselves,” said Shivraj Sabale, chief operating officer of Xoriant, a mid-tier information technology services firm. 

Work will merge into two primary roles: One focused on product and domain knowledge, with the ability to both create and test software, and another dedicated to scaling it up for deployment. 

Three skill sets will be crucial for engineers: Deep product and domain knowledge; the ability to shift from individual architecture to broad orchestration; and the capacity to think through a project end-to-end. 

“I have spent the last 20 years being a great architect, but now I have to let an agent write the code. However I will orchestrate those agents. The ability to let go of some of that is a mindset change, not a technology challenge,” said Sabale. 

Bhadada, of Neovay, said GenAI will turn the classical talent pyramid in software development into a barbell structure. There will be entry-level workers armed with AI training, a lean middle tier, and a wider top tier composed of experts who bring architectural and design depth to engineering, alongside a thoughtful approach to the economics of tokens. “This AI moment also carries real net new opportunity to work at the level of scope and opportunity that was not possible earlier.” 

But Raghu Pareddy, founder and CEO of Wissen Technology, is not entirely certain that role compressions will happen across projects. “It can be done in a brownfield product where there’s a lot of routine maintenance work and it can be automated because everything is pretty much cooked and ready to go,” he said. 

“But if there is a brownfield product where you’re doing complex modernisation and enhancements and it requires understanding of yesterday’s systems by today’s engineers, it is a little bit tough to compress it.”