|According to Forbes, 80 percent of enterprises have active AI in production today. / Photo by: everythingpossible via 123rf|
More and more companies around the world are putting substantial investments in artificial intelligence for their operations. Reports showed that businesses have spent an average of $36 million for these efforts, and they are planning to increase this by nearly 10 percent over the next two years.
According to Forbes, a global media company focusing on business, investing, technology, entrepreneurship, leadership, and lifestyle, Vanson Bourne’s “State of Artificial Intelligence For Enterprises” study reported that 80 percent of enterprises have active AI in production today. About 48 percent of Asia-Pacific (APAC) and 39 percent of North America and Europe enterprises have AI deployed in their operations. Globally, 42 percent of enterprises are planning to further implement the technology in their businesses. Vanson Bourne is an independent and specialist market research provider for the global technology sector.
The study also showed that the most effective revenue driver from AI today is product innovation and research and development (50 percent), followed by customer service (46 percent), supply chain and operations (42 percent), security and risk mitigation (40 percent), and sales (34 percent).
However, not all AI projects have successful outcomes.
In 2016, Microsoft launched a chatbot called Tay, which was meant to test and improve the tech giant’s understanding of conversational language. The chatbot could automatically reply to people and engage in “casual and playful conversation” on Twitter. However, online trolls “corrupted” Tay’s personality in less than 24 hours. As a result, it became a racist, misogynistic, and anti-Semitic chatbot.
IBM developed “Watson for Oncology,” which aimed to enable clinicians to “uncover valuable insights from the cancer center’s rich patient and research databases.” However, upon checking, medical experts discovered that it made several unsafe and incorrect treatment recommendations.
Lexalytics, an online site that provides sentiment and intent analysis to an array of companies using SaaS and cloud-based technology, reported that the company spent more than $62 million on the project without reaching its goals.
Many AI Projects Are Failing
In 2017, research firm Gartner reported that the estimated revenues for cognitive and AI systems could reach $12.5 billion that year, which was an increase of more than 60 percent from 2016 revenues. However, they found out that only 1 in 25 CIOs in the last quarter reported using AI at all. This came as a surprise since there has been an increased interest in and adoption of AI in enterprises.
Pactera Technology, an IT consulting and outsourcing company, recently released its “Artificial Intelligence Localization, Winners, Losers, Heroes, Spectators, and You” report. It revealed that 85 percent of AI projects are failing to deliver on their intended promises to business. This is despite the fact that 100 percent of leaders present in the tech event showed their interest in using tools like AI-powered Neural Machine Translation (NMT). However, organizations are still cautious about adopting new AI-related innovations with only 23 percent saying they currently use the technology.
The findings of the report showed that 8 out of 10 companies that utilize AI and machine learning stated their projects had stopped. Also, about 96 percent of them reported experiencing problems with data quality, data labeling, and building model confidence. According to TechCrunch, an American online publisher focusing on the tech industry, about 77 percent of those surveyed reported facing challenges to entry from senior management not seeing the value or wanting to invest in the emerging technology.
In a press release, Jose Martinez, vice president of digital innovations and solutions at Pactera, said, "Identifying business goals that AI can readily achieve, like Neural Machine Translation, and managing the teams that scrutinize data is what ultimately improve a business' leveraging of AI."
Why AI Projects Fail
One of the major reasons why an AI project fails is because businesses didn’t establish in the first place how AI applications will drive actual business results. Roman Stanek, CEO at GoodData, stated that companies should know what business outcomes they are trying to reach. Most of the time, these projects are being implemented without a thorough plan. Eventually, front-line managers and employees don’t find it useful. There are also some instances where there is a misalignment of expectations in these projects versus the reality of the project and its time frame.
According to CMSWire, an online site that covers digital customer experience, martech, digital workplace, and information management sectors, problems in AI projects arise when the data provided to AI and machine learning systems are either incomplete or unbalanced. The systems will not be able to cover all possible scenarios during the implementation or might cover all of the most likely scenarios but contain very few examples of some of the cases. As a result, AI and machine learning will not have the full capacity to develop a complete picture. Thus, they are required to undergo additional training.
|Problems in AI projects arise when the data provided to AI and machine learning systems are either incomplete or unbalanced. / Photo by: dolgachov via 123rf|
Jon Stenstrom, co-founder and CEO of Quantified Skin, stated that there could be underlying biases that are based on the developers’ own biases. “Some have shown a bias toward certain ethnicities or personality traits for them to enter the company's job screening process. If this goes unchecked, it opens the company up to not finding the best candidates for the positions but also to discrimination lawsuits,” he said.
Indeed, AI and machine learning systems should undergo extensive training. They need to learn how projects will bring success to businesses. These systems need to be trained not only on how to deal with failure but how to gracefully deal with failure. Ultimately, companies must first plan thoroughly every AI project they make to ensure that it will be a success.