AI agents set to transform software development - study

SoftServe has published a study on the use of agentic AI in software engineering. Conducted by MIT Technology Review, the research surveyed 300 senior technology executives.

The findings suggest large companies expect AI agents to play a much broader role in software development over the next two years. It found that 72% of organisations expect AI agents to manage most or all software or product lifecycles end to end within that period, while 41% think this will happen within 18 months.

Nearly all respondents expect AI agents to speed up software delivery, and 98% believe these tools will dramatically reduce the delays often associated with stalled pilot projects within the next two years.

AI use in software teams is already well established. Some 79% of respondents said they have used tools such as AI assistants during the past two years, with coding and quality assurance cited as the biggest early areas of benefit in the first year, at 44% and 38% respectively.

Adoption Plans

The research points to broader adoption across software teams. Half of organisations described agentic AI as a top investment priority for software engineering today, and 84% said it would become the leading investment area by 2029.

Current use remains relatively limited: 51% of software teams are using agentic AI today, while another 45% plan to adopt it within the next 12 months.

Expectations for the scale of improvement were more mixed than the headline adoption figures. Over the next two years, 14% of respondents said gains from agent use would be slight, while 52% expected moderate improvements. A further 32% anticipated stronger gains, and 9% said the effects would be game changing.

Skills Shift

The survey also points to changes in hiring as AI use expands. Respondents said the roles likely to be in highest demand are AI engineers, cited by 51%, followed by software architects at 32% and data engineers at 29%.

That marks a shift from the previous two years, when DevOps, cloud and full-stack engineers were given higher recruitment priority. The findings suggest employers expect to need more staff who can work directly with AI systems as those systems become more embedded in software development.

The report also highlights a gap between what executives see as the main barriers to adoption and what experts interviewed for the study emphasised. While 44% of respondents pointed to computing costs and AI integration as the biggest challenges, only 12% identified change management as a primary barrier.

That contrasts with the views of specialists cited in the report, who described organisational change as the central issue. The study suggests this mismatch may mean some companies are underestimating the difficulty of adoption and overestimating how quickly they can scale these tools across teams and processes.

Those experts argued that broader changes will be needed across software engineering teams, workflows, processes and the software development lifecycle if companies are to make wider use of AI agents.

Serge Haziyev, Chief Technology Officer, Advanced Technologies, SoftServe, said the current shift goes beyond the use of individual coding assistants.

"How AI is used in software engineering continues to accelerate and evolve from individual vibe coding to disciplined agentic engineering-already showing evidence of order of magnitude productivity gains," Haziyev said.

"The implication is profound for the industry, but the technology alone will not deliver the promised leap. From our perspective, the turning point for the next era of software development will come not just from technology but from completely rethinking the people, processes, and tools together," he added.

The study drew on responses from chief information officers, chief technology officers and chief AI officers, as well as other senior leaders in software engineering, data and technology. All respondents worked at large enterprises across seven industries and six countries, and most represented businesses with annual revenue of USD $500 million or more.

The report also included interviews with Nathaniel Barnes of Hg Capital, Rick Kazman of the University of Hawaii, Shashi Prabakhar of Amazon Web Services, Martin Vesper of Pfeifer & Langen, and Haziyev. Their comments suggest that while companies expect faster software delivery and less development friction, the broader impact of agentic AI will depend on how organisations adapt their structures and working methods.