Agentic AI governance is the gap that 94% of enterprises have acknowledged and only 12% are actually closing.New research finds that AI adoption has outpaced the controls meant to manage it, and legacy systems are making it worse.
A survey of 1,879 IT leaders released today by OutSystems paints a picture that will be familiar to anyone who has watched enterprise technology move in waves: broad adoption first, governance somewhere after. The 2026 State of AI Development report finds that 97% of organisations are already exploring agentic AI strategies, and 49% describe their capabilities as advanced or expert.
But only 36% have a centralised approach to agentic AI governance and just 12% use a centralised platform to maintain control over AI sprawl. That 82-point gap between awareness and action is the real headline from this report. Not the adoption numbers, which are impressive on paper, but what’s underneath them.
The agentic AI governance gap enterprises can’t close
Agentic AI, unlike its generative predecessor, doesn’t wait for a prompt. It plans its own steps, calls APIs, monitors outcomes, and operates in the background without constant human input. That autonomy is the point and it’s also why governance is harder to retrofit once agents are running.
The report makes this tension explicit. According to OutSystems, 94% of organisations say AI sprawl is increasing complexity, technical debt, and security risk. Only 39% describe themselves as “very” or “extremely” concerned. And only 12% have done something about it.
The companies managing this best are treating governance as an architectural decision, not a compliance checkbox. The report notes that 66% of leaders find building human-in-the-loop checkpoints technically difficult, requiring a complex orchestration layer to pause agents, preserve reasoning context, and produce decision logs for human review.
Most are settling for a more passive “human-on-the-loop” model instead, which works only when governance structures are already in place. For most organisations, they are not. Some 41% rely on project-level rules rather than any centralised framework, leaving compliance gaps across the enterprise.
Legacy systems are the wall agents keep running into
The adoption enthusiasm runs straight into an older problem. Legacy fragmentation and integration difficulties are the top barriers to AI development success, cited by over 40% of surveyed leaders across seniority levels, from C-suite to senior managers. Some 38% of respondents list legacy systems as the primary reason agentic AI projects have stalled.
This is not a new complaint. But agentic AI makes it sharper because agents need to move across systems, not just query one. An agent handling a customer service workflow in financial services might need to touch a CRM, a core banking system, a compliance log, and an external API–all in sequence. If any of those systems can’t be reached cleanly, the agent breaks down or, worse, behaves unpredictably.
The report finds that 48% of respondents identify integration with legacy systems as the single most important capability they need to expand agentic use. Building agents that work around current complexity–rather than waiting for legacy modernisation–is the practical answer most enterprises are gravitating toward.
Where the ROI is actually landing
Despite the governance gap, the returns are real in specific areas. Some 40% of realised ROI is coming from IT development and productivity; agents assisting with code generation, testing, and workflow automation. Operational efficiency follows at 22%. Customer experience is at 14%, suggesting that the most visible consumer-facing applications are still building trust before enterprises will let agents run those interactions autonomously.
The trust data bears this out. Overall, 73% of respondents express high or moderate trust in agents acting autonomously, up significantly from last year, when only 40% said they mostly trusted generative AI to write code without human oversight. That shift in one year is notable. Financial services leads on complete trust in autonomous agents, at 12%–the highest of any sector surveyed.
India leads all surveyed countries in IT development and productivity ROI realisation at 50%, with Japan leading in operational efficiency at 37%. The report’s APAC findings reflect a region that has moved faster from pilot to production than most, with Brazil and India posting the highest rates of agentic projects reaching full enterprise deployment.
What the CIO mandate actually looks like
The governance, trust, and architecture models that IT leaders establish now will become the template for the rest of the enterprise. That’s not a prediction from the report, it’s an observation about how technology adoption has always worked. The systems and rules IT builds to manage its own agents become the precedent for every other department that wants to deploy them.
OutSystems CIO Tiago Azevedo outlines a five-step framework in the report: establish portfolio visibility and governance first; identify specific use cases and test before scaling; build agents that navigate current data complexity rather than waiting for clean infrastructure; keep humans accountable for agent decisions; and bake access controls and role-based permissions into the platform from the start.
The most instructive finding in the report may be the simplest: 96% of leaders say a unified platform to build, manage, and govern apps, data, and AI agents is very or somewhat important. Only 7% have actually adopted one. Enterprises know what the architecture needs to look like.
Getting there, through the fragmentation, the legacy debt, and the ad hoc governance that accumulated during the pilot phase, is the operational work ahead.
Source: OutSystems 2026 State of AI Development report, surveying 1,879 IT leaders across APAC, EMEA, AMER, and LATAM. Released April 13, 2026.
Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and co-located with other leading technology events. Click here for more information.
Tech HQ is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.
Dashveenjit is an experienced tech and business journalist with a determination to find and produce stories for online and print daily. She is also an experienced parliament reporter with occasional pursuits in the lifestyle and art industries.