Agentic AI: sprawl worries
Research finds agentic AI is becoming mainstream in enterprises but organizations fear uncontrolled growth — OutSystems reported 94% of surveyed orgs worry about agent sprawl (prnewswire.com). Reporting also indicates vendors are building background and always‑on agent features into products, increasing the need for orchestration and governance (uctoday.com) (prnewswire.com).
Companies are rolling out artificial intelligence agents fast, but many say they do not have a firm grip on how many are spreading across the business. OutSystems said 94% of surveyed organizations are worried about “AI sprawl.” (prnewswire.com) OutSystems released the finding on April 13, 2026, in its State of AI Development report. The company said it surveyed 1,900 global information technology leaders, and found 96% already use AI agents in some capacity while 97% are exploring company-wide agentic AI strategies. (prnewswire.com) An AI agent is software that does more than answer a prompt once. OutSystems and MIT Sloan Management Review both describe agentic systems as tools that can plan, take multistep actions, and adapt as they go, which puts them closer to workflow software than to a chatbot. (prnewswire.com) (sloanreview.mit.edu) The worry is not only that agents are arriving, but that they are arriving in pieces. OutSystems said most enterprises still lack a centralized governance model even as they add agents across fragmented environments, raising security risk, technical debt, and operational complexity. (prnewswire.com) Software vendors are also pushing agents beyond the chat box and into the background. Microsoft said in March 2025 that autonomous agents in Copilot Studio became generally available, letting companies set triggers so agents can monitor events and execute actions without a person checking in each time. (microsoft.com) Microsoft said those agents can “run in the background” while administrators keep visibility through an activity log. That design turns agents into always-on workers inside business systems, which makes inventory, permissions, and audit trails harder to manage when every team can build its own. (microsoft.com) The same shift is visible outside Microsoft. OpenAI introduced Operator on January 23, 2025, as a service that can use a browser to carry out tasks on a user’s behalf, and Anthropic now markets Claude Code for Enterprise as an autonomous coding agent that companies can deploy and manage centrally. (openai.com) (claude.com) Researchers have been warning that adoption is outrunning management. MIT Sloan Management Review and Boston Consulting Group wrote in November 2025 that agentic AI is spreading faster than leaders can redesign processes, assign decision rights, or rethink accountability, and 76% of executives in their survey said they see agentic AI more as a coworker than a tool. (sloanreview.mit.edu) OutSystems said financial services and technology companies report the highest production deployment, and Gartner predicts 40% of enterprise applications will include task-specific AI agents by the end of 2026. The immediate problem for large companies is no longer whether agents will show up, but whether anyone can still count, govern, and shut them off. (prnewswire.com)
