In collaboration withCeligo
The transformative capabilities of AI are clearly recognized. Enterprise applications are gaining traction, and organizations are evolving beyond pilot schemes to actual AI deployment. Businesses are not merely discussing AI; they are reallocating budgets and resources to realize it. Many are currently testing agentic AI, which offers unprecedented levels of automation. However, the path to full operational effectiveness is still ambiguous for numerous entities. While AI experimentation is prevalent, comprehensive enterprise adoption continues to be a challenge.
Without cohesive data systems, reliable automated workflows, and structured governance, AI initiatives may remain in pilot stages and face difficulties transitioning to full production. The emergence of agentic AI and heightened model autonomy emphasize the necessity for a comprehensive strategy in integrating data, applications, and systems. Lacking this, AI initiatives in enterprises could falter. Gartner anticipates that over 40% of agentic AI projects may be abandoned by 2027 due to expenses, inaccuracies, and governance issues. The core problem is not the AI technology itself, but rather the lack of an operational infrastructure.
To gain insights into how organizations are organizing their AI operations and how they are successfully executing AI projects, MIT Technology Review Insights conducted a survey of 500 senior IT leaders from medium to large companies in the US, all of which are engaging with AI in some capacity.
The findings from the survey, along with a series of expert discussions held in December 2025, reveal that a solid integration foundation corresponds with more sophisticated AI implementations, beneficial for enterprise-wide projects. As AI technologies and applications advance and increase, an integration platform can assist organizations in preventing redundancy and siloed operations, while providing clear supervision as they manage the expanding autonomy of workflows.
Key insights from the report highlight the following:
Some organizations are advancing with AI. In recent times, numerous studies have highlighted a deficit of concrete AI success. Yet, our research indicates that three out of four (76%) surveyed companies have at least one department with an AI workflow fully operational.
AI is most successful when applied to well-defined, established processes. Almost half (43%) of organizations are achieving success with AI applications linked to clear and automated processes. A quarter are thriving with new processes. Additionally, one-third (32%) are integrating AI into various processes.
Two-thirds of organizations do not have dedicated AI teams. Only one in three (34%) organizations maintain a team specifically focused on managing AI workflows. One in five (21%) indicate that central IT handles ongoing AI maintenance, while 25% designate the responsibility to departmental operations. For 19% of organizations, the duties are distributed among various teams.
Enterprise-wide integration platforms foster more effective AI implementation. Organizations utilizing enterprise-wide integration platforms are five times more likely to incorporate a wider array of data sources in their AI workflows. Six in ten (59%) rely on five or more data sources, compared to merely 11% of organizations utilizing integration for particular workflows, or 0% among those without an integration platform. Those employing integration platforms also exhibit increased multi-departmental AI implementation, greater autonomy in AI processes, and heightened confidence in delegating autonomy in the future.
This material was produced by Insights, the custom content division of MIT Technology Review. It was not composed by the editorial staff of MIT Technology Review. It was researched, crafted, and developed by human writers, editors, analysts, and illustrators. This encompasses the creation of surveys and the gathering of data for those surveys. AI tools that may have been utilized were confined to secondary production processes that underwent thorough human evaluation.
