A global survey of 628 enterprise IT leaders conducted by the Futurum Group finds that increasing investment in generative artificial intelligence (AI) (40%), followed closely by AI and machine learning (ML) technologies (39%) and applying AI to IT operations (37%), also known as AIOps, are expected to be the top three drivers for accelerating software delivery over the next 12 to 18 months.

As such, the survey finds the top areas of investment over the same period are AI Copilot/AI code tools (38%), AI agent development (37%), AI-assisted testing (37%) followed closely by DevOps (37%), automated deployment (34%), software security testing (31%).

Overall, a full 60% of respondents said their organization is now actively using AI to build and deploy software, the survey finds.

Mitch Ashley, vice president and practice lead for software lifecycle engineering at Futurum Group, said it’s clear organizations are seeing productivity benefits that enable them to increase the velocity at which software is developed without increasing headcount. For example, less than a quarter of respondents (22%) said they are prioritizing investments in additional IT headcount over the next 12 to 18 months.

Less clear, however, is the degree to which the pace at which software is being built and deployed has improved. Only 12% of respondents are now updating code bases daily, compared to 37% that update weekly and 30% that update monthly. The issue holding organizations back today from adopting AI more broadly is operational friction and DevOps maturity, said Ashley.

Weekly and monthly release cycles persist because cloud-native complexity, testing bottlenecks, and operational load overwhelm human-scale processes, he added. Rising investments are being made in AI agents, AIOps, and automated security to turn AI into a workforce multiplier rather than a productivity sidecar, said Ashley.

In general, DevOps teams are trying to strike a balance between enabling developers to write more code faster and quality assurance. Much of the code created by the first wave of AI coding tools contains vulnerabilities and tends to be overly verbose, which makes it more expensive to run in a production environment. As a result, many application developers instead of writing code are spending more time reviewing code before it is checked into a DevOps pipeline.

In the meantime, however, it’s also clear that the overall size of the code base that needs to be managed by DevOps teams is steadily rising, which will require many DevOps teams to revisit how their pipelines are currently constructed and possibly the platforms and tools they are relying on today to manage DevOps workflows.

The one thing that is certain is that the output of AI coding tools is only going to continue to improve as more reasoning capabilities are embedded into large language models (LLMs). Before too long, most DevOps engineers will be invoking multiple AI agents, which in turn will be interacting with other AI agents that are performing tasks on behalf of the entire team. The challenge now is putting the orchestration frameworks that will be needed to manage interactions between human engineers and AI agents and the AI agents themselves to ensure software is actually being built as intended.