€720m in Failed AI Projects Should Be a Wake Up Call for Irish Enterprise

Irish organisations are at a critical turning point.

Our latest research reveals that €720 million has been spent on AI initiatives that failed to deliver meaningful outcomes. Even more striking, 99% of large organisations in Ireland have experienced at least one failed AI project, with the average loss per company reaching €770,000.

That data points to a difficult but necessary truth: AI is not failing businesses. Businesses are failing AI.

In practice, this isn’t about access to technology. Most organisations already have that. The issue is how decisions are being made. AI ambition is outpacing business discipline. Too often, organisations jump to “have you used AI for that” before establishing where it actually makes sense, what outcome is expected, or what risks are being introduced.

The consequence is not just wasted spend. It is exposure. Poorly governed AI creates risks to brand, regulatory standing, and long term competitiveness.

Our research highlights a clear disconnect. While 62% of companies see AI as critical, only half have a defined strategy. More concerning still, over half of organisations that have deployed AI have experienced situations where they could not explain the outputs produced by their systems.

At that point, the issue is no longer technical. It is about trust. When AI becomes a black box, organisations lose confidence internally, credibility externally, and the opportunity to be among the small number of programmes that actually deliver value.

The root cause is not a lack of innovation. It is a lack of governance.

AI is still too often treated as an experiment or an innovation initiative, which lowers scrutiny and allows ambiguity to persist. In reality, it should be approached with the same discipline as any other investment. That means clear ownership, defined outcomes, and an understanding of what success and failure look like before work begins.

This is where most organisations struggle, and where we focus our work at Saros Consulting. Not on accelerating AI adoption, but on ensuring it delivers value without creating risk.

Our approach centres on establishing a clear governance foundation through a hub and spoke model, where a central AI charter and control framework sets the standards, and business units innovate within them. This allows organisations to move quickly, but with clarity, accountability, and control.

Frameworks such as ISO 42001 and the EU AI Act provide a practical foundation for doing this properly, from data quality and risk management through to human oversight and accountability. The regulatory environment is also changing rapidly, with non compliance carrying material financial and operational consequences.

The difference between success and failure is already visible. McDonald’s withdrew a high profile AI initiative after real world failures exposed weaknesses in oversight and testing. Wendy’s, by contrast, succeeded by positioning AI as an assistant, supported by strong governance and direct leadership involvement.

The lesson is clear. AI success is not driven by the technology itself, but by how it is governed.

For technology leaders, the question is not whether to invest in AI, but whether the organisation is equipped to do so responsibly. That means having clarity on accountability at executive level, understanding where AI is being used across the organisation, ensuring human oversight is built in, and demanding transparency from vendors and internal teams alike.

The €720 million figure should not be seen as a warning alone. It is a reflection of what happens when organisations move faster than their ability to govern. The “wild west” phase of AI is already proving costly.

The organisations that succeed will not be those that adopt AI the fastest, but those that apply it with the greatest discipline. Governance is not a constraint. It is the foundation for turning AI risk into real, sustainable advantage, and for outperforming those still treating AI as an experiment.