Future Software Engineering: Redefined by Innovation

The 'Agentic' capabilities of Artificial Intelligence (AI) are heralding a third revolution in software engineering, redefining future development paradigms. Software engineering has already experienced two major shifts in the 21st century: the rise of the open-source movement and the adoption of DevOps and Agile methodologies. This article examines the emergence of agentic AI within this historical context, aiming to explain how it will transform the entire software development process.

Agentic AI has the potential to automate and manage significant portions of entire software projects, acting as autonomous agents that can reason and set direction, going beyond merely assisting with individual tasks like coding or testing. This development promises end-to-end software process automation, along with agent-driven development and product lifecycle automation.

A survey of 300 engineering and technology executives indicates that software engineering teams recognize the potential of agentic AI and have begun adopting it, albeit currently on a limited scale. Nearly half of all organizations view agentic AI as a top investment priority for software engineering, and more than four out of five expect it to be a major investment area within two years. This investment is driving accelerated adoption: 51% of software teams are currently using agentic AI, with another 45% planning to implement it within the next 12 months.

Over the next two years, most teams anticipate only minor to moderate improvements from agent use, though about a third have higher expectations. The primary benefit cited for using agentic AI is increased speed, with nearly all respondents (98%) expecting accelerated delivery of software projects from pilot to production, an average increase of 37%. Teams looking to scale agentic AI want AI agents to manage the product development and software development lifecycle (PDLC and SDLC) from start to finish, a goal that 41% of organizations aim to achieve for most or all products within 18 months.

However, at this early stage, computing costs and integration with existing systems are identified as major challenges. For all survey respondents, particularly early adopters, the challenges of integrating agents with existing applications and the cost of computing resources emerged as the most significant hurdles when applying agentic AI to software engineering. Experts interviewed emphasized the greater importance of change management.