The role of AI in software development
The transition to the product development lifecycle (PDLC) signifies a bold departure from conventional methods. The AI orchestrator is emerging as the central driver of this transformation, assuming the role of an intelligent manager capable of ingesting a system description provided by a technical product manager or architect. Rather than relying on disparate manual efforts, the orchestrator interprets this high-level, human-readable intent — what we call the “language of intent” — and seamlessly coordinates every aspect of product creation.
For example, suppose the objective is to develop a customer relationship management solution for an oil and gas client, complete with pipeline failure detection based on telemetry data. Traditionally, this would require software developers to manually code the solution, integrate various modules, craft end-user documentation, prepare marketing materials and configure the operating environment on a platform such as Microsoft Azure. Each step would often involve different specialists and significant coordination overhead.
The AI orchestrator, by contrast, automates all parts of this process. Upon receiving the human narrative description of the desired system, the orchestrator will first have a back-and-forth interaction with a small group of people, usually experts in different areas, who author the narrative to resolve ambiguities, extract more detailed insights and arrive at informed decisions. The orchestrator then autonomously generates the underlying software, produces comprehensive end-user documentation, creates training videos and even drafts marketing content — without the need for manual intervention at each stage.
This intelligent orchestration means that a fully realized product is delivered from a single, unified intent. The core distinction between the PDLC and the traditional SDLC lies in this expansive scope: The PDLC, powered by the AI orchestrator, focuses on delivering a complete, market-ready product, whereas the SDLC has historically concentrated only on producing the software component of that product.
This addresses the first pain point in the software development lifecycle. The second pain point has to do with moving forward with the underlying software tooling that’s required, at least in the past, to bring a software product to market. Tooling encompasses the full menu of software utilities: integrated development environments, version control systems, testing frameworks and monitoring systems, among others that enable the product to be used.
A software developer using the traditional SDLC, even an agile version, would still need to use a tremendous amount of tooling to write the code, test the code, push the code to development environments and test environments, and then document all product requirements. Over the years, a tremendous ecosystem has developed to support bringing a software product to market.
Implications of AI-driven development
AI-driven software development dramatically accelerates the pace of product creation while significantly reducing the resources required. Rather than assembling large teams of developers, product managers, marketers and technical writers, organizations now need to identify someone who can articulate the desired system in clear human language. The AI orchestrator transforms this input into a comprehensive, market-ready product, achieving remarkable results with a fraction of the traditional resources. Organizations that can make this shift find it to be a powerful value driver, enabling them to achieve robust outcomes, with far greater efficiency.
The implications extend far beyond incremental improvement. While current approaches often treat AI as a bolt-on feature to optimize specific aspects of the SDLC — yielding modest gains in speed or automation — the true potential of AI lies in fundamentally reimagining the process. The PDLC model abandons the constraints of the traditional SDLC altogether. By blowing up the traditional SDLC, the PDLC leverages AI to reinvent how products are conceived, built and launched, which signals a major transformation in both workflow and organizational structure.
We used the PDLC to create a demo for a proprietary product that we’re planning to bring to market as an enterprise-grade investment management tool that offers deep research on specific stocks and identifies those to blacklist, which is an important consideration for EY teams as part of our independence obligations.
We fed all the product documentation into a set of tools in an AI-centric development environment, which then created the software product along with full end-user documentation. The platform provides individual investors with personalized investment insights and strategies, leveraging advanced algorithms and data analytics. Without using PDLC, this would have taken a team of six or seven developers a minimum of 10 weeks to produce. PDLC created the investor buddy in two days, developing an enterprise-grade platform featuring authentication and authorization systems and a system that enabled back-end resources to monitor the ongoing use of the product.
With PDLC, it is now possible to achieve all three legs of the iron triangle.2 This concept essentially states that you can choose between performance, cost and quality but only achieve two of the three. For example, you can develop a product of high quality quickly, but it will be costly. The PDLC approach actually enables you to choose all three: obtain a better-quality outcome, a product that delivers higher performance in less time and at a significantly reduced cost. That’s an important benefit because achieving all three has always been viewed as untenable in the past.
Transformation of roles in software development
As organizations begin to reap the full benefits of adopting the PDLC model, the roles of software developers and product managers will transform significantly. Instead of being sidelined, these professionals will need to evolve their mindsets to take on new responsibilities that align with an AI-driven development landscape.
For example, they will no longer need to oversee traditional coding or project management tasks, which we refer to as “toil tasks.” These are tasks where, instead of using their creative minds, software developers must document things painstakingly and check all required boxes. Few people enjoy these jobs, and PDLC, along with AI in general, will eliminate the need to perform these toil tasks altogether.
We are likely to witness a grand repurposing in this next wave of software development. While some roles may become redundant, this shift doesn’t mean that people will lose their jobs. Instead, their focus will evolve as they help organizations address the backlog of software development features and projects that have accumulated over the past several years due to resource constraints, shifting business priorities and the increasing complexity of technological initiatives.
This presents an opportunity for organizations to upskill these employees and help them become managers of creativity through AI, leading their AI teams to produce superhuman results. This advancement will enable both EY teams and our clients to tackle these backlogs far more efficiently.
Rather than diminishing the importance of software developers or product managers, this evolution redefines their roles. Moreover, this will likely lead to a surge in new developers and a proliferation of innovative applications as more individuals harness the power of AI-driven tools. Many organizations can tap into the expertise of these evolving professionals to aggressively address expanding backlogs, which have only continued to increase in recent years, and meet the accelerating demand for new products and features.
By empowering developers and managers to lead teams of AI agents, organizations are equipped to tackle pressing needs with greater speed and efficiency. The evolving roles foster creativity, strategic oversight, and the ability to drive transformation within the organization, demonstrating that human expertise remains essential, even as technology advances.
Five years from now, we could very well begin to see many larger, serious enterprise systems up and running in production that are entirely developed by a couple of human actors, running with AI at the core. The AI orchestrator will perform the bulk of the toil tasks, while the people perform intent matching, verifying that the system output matches the language of intent in a meaningful way.
As Professor Alex Pentland of MIT notes, “Instead of killing jobs, AI will create new more valuable ones, but people need to be ready to step up to the challenge and learn the new skills. Those who remain tied to the past ways of working will not flourish.”
Challenges and considerations
As organizations embrace AI-powered tools, they must carefully consider the potential impacts, especially when these systems begin to take on tasks previously managed by humans. Issues such as algorithmic bias, transparency in decision-making and accountability for outcomes are critical. As AI increasingly influences how products are conceived and delivered, it is essential to establish frameworks that uphold fairness, protect user privacy and enable responsible use of technology.
The need for human oversight and intervention
Despite the advanced capabilities of AI, human oversight remains indispensable. AI systems, regardless of their sophistication, can make errors or produce unintended consequences. Organizations need to take steps to maintain a strong role in human judgment, so AI agents are monitored and guided by experienced professionals. Overreliance on automation without adequate checks can lead to ethical lapses or operational failures. Therefore, embedding oversight mechanisms and fostering a culture of accountability are necessary to balance innovation with responsibility.
