It’s the night before a major client meeting, and a new kind of digital team is working hard behind the scenes. But this isn’t a human team — it’s a team of fully autonomous AI agents interacting to perform complex tasks with minimal human intervention. One AI agent scans thousands of pages from internal knowledge bases, another pulls recent industry reports, while a third combs through 10-K filings from the client and its competitors. Yet another agent distills all these insights to build a briefing deck. By the time the consulting team logs on the next morning, everything they need — summaries, key talking points and data visualizations — is waiting in their inbox.
Scenarios like this aren’t confined to internal processes. Across all industries, agentic AI is poised to transform myriad business processes from customer service to business operations and software development. For developers, it’s already redefining how teams prepare, build, and deliver their software products, either for internal or external customers. Nearly half of organizations are deploying agentic orchestration to power code reviews, automate testing, and accelerate deal cycles, making once-impossible productivity gains a daily reality.
So how did we get here? To better understand the growing impact of agentic AI, OutSystems partnered with KPMG and CIO Dive to survey 550 software executives and find out where they are in their agentic AI journeys and the risks and opportunities that face them.
From pilots to critical mass
Just a year ago, software teams were heavily focused on generative AI for things such as code suggestions and bug identification. Now, these tools have become everyday teammates, with AI moving from pilot projects to near-universal adoption. Moreover, almost all organizations we surveyed said they planned to increase their investment in AI over the next 12 months — with agentic tools now taking center stage. The latest research found that 47% of organizations were actively integrating agentic systems into their workflows, while almost all others were either running pilot projects or planning to within the next few months.
According to the survey, AI — traditional or agentic — was making the biggest mark in testing and quality assurance, development and coding, and monitoring and maintenance. However, and partially thanks to agentic systems, adoption was quickly spreading into other areas of the software development life cycle, such as requirements gathering, design and integration. Cloud-based AI services were fueling much of this momentum, but a significant portion of organizations were also building their own AI agents in-house using open-source frameworks and low-code platforms.
Clearly, the scope of AI adoption has gone far beyond automating routine tasks. With agentic AI presenting the next sea change across the industry, development teams are now integrating these tools to perform complex tasks so they can focus on big-picture problem solving. Software executives were already reporting measurable impacts, too, including increased developer productivity, improved software quality and improved scalability of development activities. In other words, developers are getting more done in less time and with fewer errors.
Scaling agentic AI with trust, not chaos
There’s no doubt about it — AI is redefining what’s possible. However, the rise of agentic AI doesn’t mean developers are obsolete. In fact, scaling with trust requires that entirely new roles emerge — such as agent architects, prompt engineers and orchestration leads — with humans that bring creativity, empathy and ethical oversight. After all, rapidly adopting any new technology also brings new complexities. With new AI tools and services entering the market almost daily, the risk of tech sprawl and the unsanctioned use of AI (shadow AI) is all too real. Software executives also reported other concerns, particularly in areas such as security and compliance, data quality and reliability, and explainability and accountability.
Agentic AI is about automation using agents that act instead of waiting for instructions. That requires oversight, which, in turn, requires consolidation and standardization, especially since managing and monitoring a fragmented tech stack is far more complex. The challenges of tech sprawl, integration with legacy systems and maintaining transparency across AI-driven decision-making will only intensify as adoption increases. The human element — reskilling, onboarding new roles, and working closely with consultants or external vendors — is more important than ever.
To address these challenges and safely deploy governable AI technologies across their software development, 46% of software executives said they relied on low-code platforms. The platform approach is fast emerging as the most viable strategy to unify AI tools and workflows in a single, secure environment that allows organizations to orchestrate both human and AI efforts with clarity, speed and control. The future belongs to those who scale with intention, keeping humans in the loop and making AI work for people — not the other way around.
Get the full report for complete access to the 2025 OutSystems and CIO Dive survey. ↗
