Irish businesses have collectively lost €720 million on failed artificial intelligence projects according to a March 2025 survey, highlighting the significant financial risks companies face as they attempt digital transformation amid escalating technology costs. The substantial losses underscore growing concerns that returns from AI investments are failing to justify their increasingly expensive implementation and operational requirements.
The emergence of what industry observers term ‘token thrifting’ represents a strategic shift in how Irish companies approach AI deployment. This cost-conscious methodology contrasts sharply with earlier ‘token maxxing’ approaches where businesses pursued maximum AI capabilities regardless of expense. Enterprise Ireland supported companies are now reassessing their technology strategies to ensure artificial intelligence investments deliver measurable business value rather than pursuing innovation for its own sake.
Token costs, which represent the computational units required to process AI queries and generate responses, have become a critical budget consideration for Irish businesses. Companies using large language models and generative AI tools face mounting expenses as these systems consume significant processing power for each interaction. A single complex query can cost multiples of what simpler requests demand, creating unpredictable cost structures that challenge traditional IT budgeting practices.
The €720 million figure represents a substantial drain on Irish business resources at a time when economic uncertainty demands prudent financial management. These failed projects encompass initiatives that were abandoned mid-development, implementations that failed to deliver promised benefits, and systems deployed but subsequently discontinued due to poor performance or unsustainable operational costs. The Central Bank of Ireland has noted increased scrutiny of technology spending in financial services as institutions balance innovation requirements against risk management protocols.
Irish organisations are now implementing more rigorous evaluation frameworks before committing to AI projects. This includes detailed cost-benefit analysis, proof-of-concept trials with limited scope, and clearer definition of success metrics before full-scale deployment. Technology leaders are prioritising use cases with direct revenue impact or measurable efficiency gains over experimental applications that lack clear business justification.
The shift toward token thrifting involves several practical strategies. Businesses are optimising prompt engineering to reduce unnecessary token consumption, implementing caching mechanisms to avoid repeated processing of similar queries, and selecting appropriately sized AI models rather than defaulting to the largest available options. Some companies are exploring open-source alternatives to proprietary AI services, reducing licensing costs while maintaining functionality.
Financial services firms operating in the IFSC have proven particularly sensitive to AI cost management given their regulatory obligations and established risk frameworks. These organisations require robust governance around AI deployments, including audit trails, explainability requirements, and compliance documentation that can add significantly to implementation complexity and expense. The IDA Ireland has emphasised the importance of sustainable technology adoption for multinational operations based in Ireland.
The failed projects data reveals common patterns across sectors. Many initiatives suffered from inadequate data preparation, with companies underestimating the quality and quantity of information required for effective AI training. Others encountered integration challenges connecting AI systems to legacy business processes, while some faced user adoption resistance when deployed tools failed to integrate smoothly into existing workflows.
Industry analysts suggest the €720 million loss figure may actually understate the true cost when accounting for opportunity costs, diverted internal resources, and delayed alternative technology investments. The learning curve associated with artificial intelligence has proven steeper than many Irish businesses anticipated, with technical expertise shortages contributing to project difficulties.
Looking forward, Irish businesses are adopting more measured approaches to AI investment. This includes phased rollouts with defined checkpoints for evaluation, greater emphasis on vendor partnerships with clear accountability structures, and increased investment in staff training to build internal AI literacy. The focus has shifted from maximising AI capabilities to optimising AI efficiency, ensuring technology spending delivers tangible business outcomes rather than simply demonstrating innovation credentials.
