Why So Many AI Projects Fail (And How to Make Yours Succeed)

Business leaders are increasingly impatient to deploy artificial intelligence (AI) in their operations, with many having high expectations for what the technology can deliver.

Tech leaders are willing to spend to reap what they hope will be game-changing business improvements and streamlined operations, with a 61% rise in planned spending on AI in 2024, according to new research[1]. But business leaders should strike a balance between their excitement for AI with the needs of the business. For all the promise of this technology, many companies have already ended up with AI proof-of-concepts which have not delivered the results they hoped for. Attaining tangible results from AI investment requires both careful thought, and attention to detail in execution.

With the huge hype swirling around AI technology over the past 18 months, business leaders have been tempted to call their IT teams and demand, ‘Why are we not using generative AI right now?’ But the truth is that often, in those businesses, both the leaders and their teams don’t actually know how to gain an advantage from AI. Leaders need to ensure that AI is rolled out for the right reasons, not just because their competitors are doing it.

There is a huge gap between exciting tech built in the laboratory and the day-to-day reality of business applications. It’s all too easy to take a short-sighted view and become over-excited by technology that has not yet crossed this gap. This is exactly how AI investment is wasted.

Why AI can fail

Even the very best technology is just a science experiment if it cannot be adopted and used in the real world. The single biggest reason AI ‘doesn’t work’ for businesses is that people try to ‘do AI’ rather than identifying where problems or inefficiencies exist. To find such problems, business leaders should first talk to partners, and listen to consumers and front-line employees. Does the business lack staff to talk to customers? Does the business need to find a way to cut fuel emissions? Beyond the hype, the real excitement of this technology comes not from thinking about AI as a standalone solution, but by adding AI into the solution to a real business problem.

How to make AI succeed

All too often, the approach to AI is to have a specific ‘AI team’, rather than applying the technology across the whole business. This siloed approach is a key mistake. AI must be integrated with a holistic approach, and a view to scaling it across every part of the business. Business leaders must connect multiple teams together to initially implement the technology, and avoid cutting corners to ensure seamless integration.

Business leaders need to design an effective proof-of-concept solution that includes AI appropriately in order to mitigate a business problem, and then scale it accordingly. For example, a generative AI chatbot that can answer niche questions could be made available to a small subset of customers initially, but rolled out to larger groups thereafter. Internal communication is also key as the business benefits of the proof-of-concept must be effectively communicated within the organisation, as AI projects often fail to be exciting to leadership until they grow to a certain size.

Why generative AI can cause problems

Even experts who have worked in the field for many years were caught by surprise at how the launch of ChatGPT made the pinnacle of AI technology so easy to adopt. This, in turn, made it easy for business leaders to imagine that generative AI should be adopted universally. But they should pause to think about whether such technology is the right choice, or if other forms of AI might do the job better.

The enthusiasm around generative AI has meant that it’s sometimes used in areas which don’t play to its natural strengths. Generative AI is great for conversational user interfaces such as chatbots, knowledge discovery and content generation. It’s also highly useful in segmentation and intelligent automation and anomaly detection. For example, Smartia, a leading UK Industrial AI & IoT technology company, worked with Lenovo to harness machine learning and computer vision AI technologies to enable its composite manufacturing process to be smoother and greatly reduce anomalies. This demonstrates how AI is already improving manufacturing quality control through various systems that accurately detect defects.

How companies ARE making AI work

Artificial intelligence is already helping organisations to solve real problems in sectors such as retail and manufacturing. AI helps to streamline and speed up processes, eliminating the amount of time spent by employees on mundane tasks. In both retail and manufacturing, computer vision is emerging as an interesting and successful use of AI, linking the physical and digital worlds, and helping to spot defects on production lines and offering valuable insight in retail settings.  

Signatrix‘s AI solution uses computer vision to draw important insights from cameras in retail stores, far beyond simply dealing with theft or similar incidents. The system is able to offer insights into important trends around what customers are looking at and buying, and to validate the success of promotions. The system can identify everything from misplaced products to how retail media (advertising) within the store is performing in terms of views.

In manufacturing, Graymatics’ LabVista software uses computer vision to help make factories and laboratories more efficient and also safer for employees. LabVista conducts quality control checks on products, ensuring they are not missing any components, and monitors the number of products coming off a production line in any time period, also scanning for defects. But even more importantly, the LabVista system helps to make factories safer: the system scans for smoke and fire, while also detecting accident-prone machinery.

Paving the way for an AI future

Business leaders need to keep their feet on the ground when dealing with AI, and avoid being swept up in the excitement around the technology. That means taking a step back and focusing on real, tangible problems and prioritising which of those to fix first. A holistic approach is key here: your AI integration should be weaved into the solution to a real-life business problem, and the project should be something that multiple teams can get ‘hands on’ with. By taking a measured, holistic approach, leaders can ensure that AI projects make it beyond the drawing board, and truly reap the rewards of this exciting technology.


About the Author

Nicholas Borsotto is WW AI Business Lead and Head of Lenovo AI Innovators Program at Lenovo. Lenovo is a US$57 billion revenue global technology powerhouse, ranked #217 in the Fortune Global 500, and serving millions of customers every day in 180 markets. Focused on a bold vision to deliver Smarter Technology for All, Lenovo has built on its success as the world’s largest PC company with a pocket-to cloud portfolio of AI-enabled, AI-ready, and AI-optimized devices (PCs, workstations, smartphones, tablets), infrastructure (server, storage, edge, high performance computing and software defined infrastructure), software, solutions, and services.

[1] IDC research, sponsored by Lenovo: ‘CIO Playbook 2024’ 

Featured image: Adobe

more insights