Agentic AI Set to Transform Software Engineering Processes

Software engineering may be on the verge of its next transformative era, with the adoption of so-called agentic AI poised to bring fundamental change to how software is developed and managed. According to a new report based on a survey of 300 engineering and technology executives, organisations are beginning to explore the technology’s potential, though widespread implementation remains in its early stages.

Since the 2000s, software engineering has changed radically, first with the rise of open source development and later through collaborative methodologies such as DevOps and agile. Now, a third shift is emerging, as AI agents—autonomous, reasoning tools—are increasingly able to not only assist with discrete tasks but also manage entire software projects with minimal human oversight. These 'agentic' AI systems are designed to automate the end-to-end processes involved in software development, from planning and writing code to testing and deployment.

Up until now, engineering teams have primarily limited their use of AI to narrowly defined tasks such as code suggestion or automated testing. Agentic AI, however, promises more comprehensive automation by acting as self-directing project managers. If widely adopted, this technology could introduce true end-to-end automation of the software development and product management lifecycle.

The report, commissioned by Insights (the commercial content arm of MIT Technology Review), finds that 51% of surveyed teams are already experimenting with agentic AI, albeit in limited applications. Nearly half have concrete plans to invest over the next 12 months, with interest intensifying: while currently half of organisations consider agentic AI a top priority, over 80% expect it to be a leading area of investment within two years.

Initial gains from agentic AI are expected to be incremental. Most executives surveyed anticipate either slight (14%) or moderate (52%) improvements in productivity and quality within two years, while nearly a third expect more substantial change, and about 9% believe the impact will be transformative. A near-unanimous 98% expect measurable acceleration in software delivery timelines, with an average projected speed increase of 37% from pilot to production.

Organisations’ ambitions extend beyond discrete automation. Forty-one percent of surveyed teams aim for AI agents to manage the entire software or product development lifecycle for most or all products within 18 months; this share is expected to rise to 72% over the following two years if projections hold.

Despite growing enthusiasm, several challenges remain. The report identifies integration with existing software systems and increased computing costs as early obstacles—particularly for fast-moving sectors such as media, entertainment, and technology hardware. Additionally, experts stress that successful adoption of agentic AI will require challenging shifts in team structures and development workflows, echoing past disruptions triggered by the uptake of agile and DevOps.

As engineering organisations brace for these changes, most agree that the potential rewards in efficiency, delivery speed, and product quality may outweigh the growing pains involved in adopting agentic AI. The report suggests that, if these hurdles can be managed, the technology could redefine industry practices over the coming decade.

Source: technologyreview.com.