Artificial Intelligence (AI) is changing the way businesses function. However, developing and successfully implementing an AI project pose significant challenges to companies, which in turn slowing AI adoption in the organisations. A new Gartner survey found that about half (54%) of all AI projects are making the leap from pilot to production. The number is slightly higher from its 2019 research report that stated 53% of AI projects fail to move from its pilot stages.
Gartner said, despite a more flexible approach to automation, brought on by the pandemic that allowed for more practical implementation of AI within organisations, there are ample challenges that exist when it comes to the actual implementation, as the models that are never deployed, scaling operations and proving positive return on investment is where the disconnect lies.
“Scaling AI continues to be a significant challenge,” said Frances Karamouzis, distinguished VP analyst at Gartner. “Organisations still struggle to connect the algorithms they are building to a business value proposition, which makes it difficult for IT and business leadership to justify the investment it requires to operationalise models.”
One problem for scaling AI is the complex governance needed when deploying numerous AI models, and 40% of surveyed organisations reported having thousands, and sometimes hundreds of thousands, of AI models deployed. More models can create governance challenges that can make demonstrating value and Return on investment (ROI) difficult, a difficulty that was seen as a top barrier to AI adoption, it said.
The survey also found other barriers to entry, including finding the right talent. Despite other technology fields reporting a talent shortage, the survey found that a dearth of AI talent is not a significant barrier to AI adoption with 72% of executives stating that they already have or can source the needed AI professionals for their projects.
“The most successful organisations use a combination of in-house development and external hiring for AI talent. This ensures that the team renews itself continuously by learning new AI skills and techniques and considering new ideas from outside the organisation,” said Erick Brethenoux, distinguished VP analyst at Gartner.
Issues regarding security and privacy were also not seen as a top barrier to AI adoption and were cited by just 3% of those surveyed, despite 41% disclosing previous AI privacy breaches or security incidents. Though it is not the most significant barrier, AI security is still a concern, with half of the respondents noting they were worried about competitors, partners, or various third parties gaining access to sensitive information, with 49% also concerned about malicious external attacks. Interestingly, 60% of respondents who have faced an AI security or privacy event reported their data was compromised by an internal party rather than from an outside source.
“Organisations’ AI security concerns are often misplaced, given that most AI breaches are caused by insiders,” said Brethenoux. “While attack detection and prevention are important, AI security efforts should equally focus on minimising human risk.”
Not just Gartner, a September 2020 survey by MIT and Boston Consulting Group suggested that global companies have reported minimal or no impact from their AI projects, citing a number of reasons for this including a lack of focus on cultural change and training within an organisation as it adapts to new working practices, but the most important factor it stated was poor data. This encompasses everything from inadequate data architecture and discovery, to modelling, quality, and governance, it said.
Despite the failures however, the same Gartner report indicates little sign of a slowdown in AI investments, as it found that 80% of executives think automation can be applied to any business decision and if it gets embedded in digital strategies from inception. Hence the analyst firm sees many organisations plan to increase these investments in the coming months.