University of Ottawa and IVADO Release Global Policy Blueprint to Prevent 80% Failure Rate in Government AI Projects

Global experts from uOttawa and IVADO offer four critical actions for policymakers to successfully integrate AI and avoid high project failure rates.

By: AXL Media

Published: Mar 9, 2026, 11:54 AM EDT

Navigating the High Failure Rate of Public Sector AI

As governments worldwide rush to adopt artificial intelligence, a new report from the University of Ottawa’s AI + Society Initiative and IVADO warns that the path to digital transformation is fraught with systemic risks. Despite 70% of nations reporting the use of AI to improve internal processes, statistics indicate that over 80% of AI projects ultimately fail to meet their objectives. The policy brief, titled Governing with AI, argues that caution and pragmatic planning are essential to ensure that these technologies do not amplify existing bureaucratic dysfunctions. According to Dr. Florian Martin Bariteau, the director of the initiative, bottom up, problem driven planning is the only credible way to transform an administration without triggering distrust from the public.

A Measured Approach to Federal AI Deployment

The report highlights Canada’s recent efforts under Mark Carney’s government as a case study in ambitious yet controlled AI utilization. The Canadian government recently employed an AI platform to synthesize 11,000 public submissions regarding its national AI strategy, signaling a shift toward large scale federal deployment. According to Dr. Catherine Régis, a professor of law at Université de Montréal, governments should adopt a "slow and steady" approach that remains ambitious from the start. This strategy is framed not as indecision, but as a mark of responsibility to ensure that institutional capacity can keep pace with technological change.

Redesigning Services Around Human Problems

The first core recommendation of the blueprint emphasizes that AI should not be deployed as a standalone solution but as a tool to solve specific, pre identified public service issues. The authors suggest that public servants must be involved as co designers to ensure that AI tools build upon proven successes rather than replacing functioning systems with unproven algorithms. According to the policy brief, successful integration depends less on how sophisticated the AI is and more on the ability of the institution to remain accountable for the outcomes. By involving the workforce in the design phase, governments can scale up effective tools while mitigating the "black box" effect often associated with automated decision making.