Why Tech Buyers Face AI-Related Project Failures

Many U.K. businesses are struggling to get their AI projects off the ground because the technology is simply not applicable, an AI strategist claims.

New research from data management platform Qlik has found that 11% of U.K. businesses have at least 50 AI projects stuck in the planning stage. Meanwhile, 20% have had up to 50 projects progress to planning or beyond — but then had to pause or even cancel them.

“AI has the potential to impact nearly every industry and department, but it’s not universally applicable,” James Fisher, Qlik’s chief strategy officer, told TechRepublic.

“Some projects fail because of infrastructure and data issues, but in other cases, AI is simply not the right tool for the job. It’s essential for businesses to understand the problem they are trying to solve and to apply AI where it can bring the most value.”

SEE: How to Improve Your Digital Transformation Project Failure Rate

This corroborates research from Gartner published in September that found that at least 30% of generative AI projects will be abandoned after the proof-of-concept stage by the end of 2025. This is not a new notion, with TechRepublic reporting on a similar finding back in 2019.

Data governance represents a key challenge

The biggest reason for AI project failures from the new Qlik research, cited by 28% of the 250 U.K.-based C-suite executives and AI decision makers surveyed, are the challenges around data governance.

“AI projects can fail to deliver in cases where there is a lack of high-quality, structured data or where objectives are too ambiguous.” Fisher said. “For example, automating customer service interactions without sufficient human oversight, the right data needed to support it or proper testing.

“Without a solid data strategy, AI models will always struggle to deliver meaningful insights.”

Incorrectly implementing a strategy can be “disastrous,” Fisher said. For example, AI-generated code has been known to cause outages, and security leaders are considering banning the technology’s use in software development.

The Qlik study also found that 41% of U.K. senior managers lack trust in AI, which could be related to other high-profile failures of late, such as Air Canada’s chatbot giving incorrect fare policy information, resulting in legal and financial repercussions. New legislation, such as the E.U. AI Act, will only raise the costs of such errors.

SEE: Generative AI: A Source of ‘Costly Mistakes’ for Enterprise Tech Buyers

But, there are business areas where Fisher has seen AI proving useful, such as supply chain optimisation, fraud detection, and personalised marketing.

“These are use cases where AI models are fed greater volumes of high-quality data, are aligned to clear business outcomes and can produce sharper, more actionable insights,” Fisher noted.

More must-read AI coverage

Reduce potential financial losses by seeking out “plug-and-play” AI solutions, experts say

Gartner estimates that building or fine tuning a custom AI model can cost between $5 million and $20 million, plus $8,000 to $21,000 per user per year. GenAI “requires a higher tolerance for indirect, future financial investment criteria versus immediate return on investment,” which “many CFOs have not been comfortable with,” analysts wrote.

Fisher emphasised the importance of business leaders ensuring that AI will deliver a real return before making the investment, and suggests trying to find an applicable “plug-and-play” solution first.

He explained: “In an environment where CIOs are already reconsidering the cost-effectiveness of generative AI solutions, a focus on smaller, purpose-driven models and targeted applications may, in the near-term, likely prove to be a more sustainable alternative.

“The simplicity of plug-and-play solutions provides businesses with a foundation for their AI projects which can help address challenges around trust and governance by reducing risk and complexity, whilst ensuring businesses are reaping the benefits that AI can offer.”

SEE: Generative AI Projects Risk Failure Without Business Executive Understanding

He also advised to start with smaller AI projects to demonstrate proof-of-concept before scaling, and to regularly assess the ROI.

“The absolute first step is to establish a strong data foundation and have the right data governance, quality and accessibility in place,” Fisher said. “Make sure you have a clear business problem or challenge in mind that AI is addressing and set measurable outcomes to track success against. To build trust in the technology, try to encourage knowledge sharing and upskilling across the business.

“Finally, take a gradual approach to AI adoption; start with a proof of concept to validate your project before committing to bigger bets.”