Many businesses in the U.K. are hitting roadblocks with their AI projects. According to James Fisher from Qlik, only 11% of companies have 50 or more AI projects stuck in planning, and about 20% have advanced to planning or beyond, only to pause or cancel them.
Fisher points out a crucial fact: while AI can change the game for many industries, it isn’t a one-size-fits-all solution. Some projects falter due to data and infrastructure challenges, but in other cases, AI just isn’t the right fit. Businesses need to identify their specific problems and apply AI in ways that add real value.
Research from Gartner supports this, indicating that 30% of generative AI projects could be scrapped after the proof-of-concept stage by 2025. This isn’t a new concern; similar findings have been reported before.
Data governance is a significant hurdle. In a recent Qlik study of 250 C-suite executives and AI decision-makers, 28% highlighted data governance challenges as the main reason for project failures. Fisher explains that without high-quality, well-structured data, projects struggle. For instance, automating customer service without the right data or proper oversight can lead to failure.
Fisher warns against poorly implemented strategies, which can have serious consequences. AI-generated code has caused outages, and some security experts are even looking to ban its use in software development. Trust in AI systems is shaky; 41% of senior managers in the U.K. don’t fully trust the technology, likely due to past failures, like Air Canada’s chatbot fiasco.
Yet, Fisher sees areas where AI can shine, such as in supply chain optimization, fraud detection, and personalized marketing. These areas benefit from large volumes of quality data and clear business objectives, leading to actionable insights.
Experts suggest businesses should consider more straightforward AI solutions to minimize financial risks. Gartner notes that developing custom AI models can be quite expensive, running between $5 million and $20 million, plus annual user costs. Many CFOs are hesitant about the upfront costs versus potential long-term gains, especially with generative AI.
Fisher stresses the need for leaders to ensure their AI investments will yield concrete returns. He recommends starting with readily available “plug-and-play” solutions. Such options make it easier to navigate issues related to trust and governance since they involve less risk and complexity.
He advises businesses to kick off with smaller AI projects to demonstrate their potential before scaling up. Regularly reassessing ROI is crucial. “First and foremost, build a solid data foundation and ensure you have good data governance and accessibility,” Fisher says. Identify clear problems AI can solve and set measurable outcomes to track success. Encouraging knowledge sharing across teams can bolster trust in AI technologies. Finally, adopt a gradual approach to AI, beginning with proofs of concept to validate projects before going all in.
