Why most AI projects fail

Verticals determining the success of any project vary even within the same field, or even within the very same project down the timeline. This is the case especially in AI, a relatively new field fueling business worldwide and shifting the attention of investors to more technologically powered, sustainable and cost-effective solutions. AI is a business superpower on the rise; the Kingdom of Saudi Arabia hosted the Global AI Summit in October 2020, where an elaborate, market-wide AI strategy was revealed. The strategy aims to train up to 20,000 data and AI experts and set up 300 AI-focused startups. This is predicted to generate up to $20bn in investment by the year 2030.

If AI can be used to develop whole nations, it’s then a tool to shift the way the world works, and a lot of decision-makers are noticing this as more than a trend, but also as a gateway into a healthier, more sustainable future.

The enthusiasm to jump on the AI train is understandable and encouraged, but the success of AI depends on the complete universe of data being captured and analyzed through a large-scale database with continuous analysis of the convergence between predictive and real-time data. Premature failure of AI projects is a common concern, and one that is legitimate; recent numbers show 85 per cent of all AI projects fail, sometimes before they have even been initiated. Why? Well, oftentimes, these big projects are misaligned with business priorities. Especially today, a lot of businesses are invested in keeping their traditional processes going, so embarking on such an altering project is not likely to make the top priorities list.

AI is expensive. Spending on AI projects and solutions will hit the $58bn mark within the next few years. Many companies tend to take a leap into a sea of information on the matter and end up with one takeaway: AI helps cut costs. While this is true, the results take time, effort and skills. From partnering with the right solution providers to hiring the perfect candidates to administer AI projects. A lot of companies will initiate and then withdraw when the costs are fully evaluated in comparison to the results. This is why strategizing AI projects is essential to success but still does not guarantee it.

While strategy is a vertical in business generally, in AI-focused projects, it’s simply not enough. Financial strategy without expertise is like throwing money on a problem; it simply won’t work. AI is new. This means there is a lack of the proper skills in the market. For reasons like this, governments are turning to training programmes, pumping capital into developing a solid generation of skilled AI engineers, scientists, and businesspersons. This relatively new field makes investment tricky also because there isn’t enough history behind it. In a way, it’s an age of trial and error, and error often means massive losses and premature failure. This happens a lot; a lot of AI projects start big, with ambition and hope backing up high-complexity work.

This relates to the way AI is described as a risky business. It’s considered an expensive tool that is hard to measure and maintain. However, developing strategies and approaches can set companies on the right path. It must always start with a problem the business is facing, and a question; ‘Can AI solve it?’ A strategy can then be designed and set, with the proper and regular measuring of ROI. There is more to this, however. Part of why AI is often met with dissatisfaction, contributing to many businesses choosing safer options, is the ambiguity and myth surrounding it. For the longest time, it seemed like this technological leap was a thing for film and literature. When investors go into AI, they have a hard time managing and adjusting their expectations. Businesses investing massive efforts and capital in AI often focus on the technology instead of the business. AI adoption must be a step-by-step process, starting with why the business needs it, and how it can help. AI can, for example accurately predict valuable information based on data, but it cannot function as intricately as the human brain.

The way an AI solution will work for a business is through data, and the quality of data it is fed with. This poses another threat to AI projects and is considered a major contributing factor in their failure. The term ‘Big Data’ came about as tech giants started taking over the global business sphere. The concept is quite enigmatic, because ‘how big is big,’ one might ask. Well, big. AI requires a lot of data to deliver, and the more the better. If a company is small, with not much data to go from, then expectations must be scaled to that level of data availability. The data must also be relevant to the problem the AI solution is designed to solve, and oftentimes such intricacies aren’t even considered.

AI is risky, but the rewards can bring cost-cutting and long-term success to the business. It is most crucial to consider data quality and availability when AI adoption is on the table. Strategy, success and failure measurement criteria are also determining steps in the adoption process. Meanwhile, globally, the direction business is going towards is one that should solve the issue of low skill levels and expertise in the field.

Anas A. Abdul-Haiy is director and Deputy CEO of Proven Consult