Why Do Company AI Projects Stall, and What Helps Adoption?

Company AI projects often stall for organizational reasons rather than model quality, according to the research reviewed here, though no study measures how often. RAND’s interviews with 65 experienced AI practitioners identified five root causes of failure, led by stakeholders misunderstanding the problem to be solved, and also pointed to data, infrastructure and leadership. Adoption is easier when leaders define a specific problem, check the data, and give managers an active role; Gallup survey data links manager support to heavier AI use, although that is an association, not proof of cause. This article is based on RAND and Gallup sources only.

Why projects fail, according to RAND

RAND’s report on the root causes of AI project failure (Ryseff, De Bruhl and Newberry, August 2024) is based on interviews with 65 data scientists and engineers, each with at least five years of experience. RAND did not measure how often AI projects fail; its study is qualitative. It defined failure as a project “perceived to be a failure by the organization,” covering both technical and business failures.

It identified five leading root causes:

• Stakeholders misunderstand or miscommunicate what problem AI should solve, so models are optimized for the wrong metrics or do not fit the business workflow.
• The organization lacks the data needed to train an effective model.
• The organization focuses on the latest technology instead of real user problems.
• Infrastructure to manage data and deploy models is inadequate.
• The technology is applied to problems that are too difficult for AI.

In RAND’s industry interviews, 84 percent of interviewees cited one or more leadership-driven causes as the primary reason AI projects fail; this is a share of those interviewees, not of all projects or organizations. RAND cautions that most interviewees were non-managerial engineers, so the results “may be skewed toward identifying leadership failures.” The report also describes a human factor: subject-matter experts sometimes put up passive resistance because they believe the projects are meant to replace their jobs.

What the evidence says makes adoption work

• Start with a specific, lasting problem. RAND recommends choosing enduring problems and being prepared to commit a product team to a specific problem for at least a year; this is a recommendation, not a threshold shown to guarantee success.
• Put technologists and business experts together. RAND says effective interaction between them “can be the difference between success and failure.”
• Invest in data and deployment infrastructure. RAND interviewees reported that models could not be moved to production because of incompatible environments or missing automation.

The manager and communication gap

Gallup’s analysis of manager support (published November 2025, from a May 2025 survey of 19,043 U.S. employed adults; individual questions were asked of smaller subsets) found that employees who strongly agree their manager actively supports their team’s use of AI are 2.1 times as likely to use AI a few times a week or more. They were also 6.5 times as likely to strongly agree that the AI tools their organization provides are useful for their work. Only 28% of employees in organizations that Gallup identifies as implementing AI strongly agreed their manager actively supports their team’s use of it. These are associations from a survey, not proof that manager support causes higher use.

Gallup’s AI indicator page reports a separate update from May 2026, which should not be combined with the 2025 survey above. 36% of employees in AI-integrating organizations strongly agreed their manager supports their team’s use of AI, and those employees were 1.7 times as likely to use AI a few times a week or more. Only 25% of U.S. employees said their organization had communicated a clear plan for integrating AI into current practices. The same page reports that 30% of U.S. employees use AI a few times a week or more, and 15% use it daily.

A practical checklist

• Write down the business problem and the metric before choosing a tool.
• Check that the data exists and fits the new purpose.
• Define success as production use with measurable results, not a finished pilot. These are suggestions drawn from the research, not a guaranteed recipe.
• Give managers a clear role in encouraging and shaping AI use on their teams.
• Communicate the plan to employees, and involve the domain experts whose work is affected.