A recent opinion poll conducted by market researcher Gartner among the IT managers of large companies (CIOs) in Hong Kong and abroad found that 4% already implemented Artificial Intelligence (AI) and 46% plan to do so.

"Despite the tremendous interest in AI technologies, current implementations remain at a fairly low level," said Whit Andrews, vice president of research at Gartner.

"However, there is potential for strong growth as IT executives of large companies start to drive AI programs through a combination of their own development and acquisition."

According to Gartner, the "early adopters" are faced with many obstacles to the use of AI. For them, the Gartner analysts have summarized four lessons from early AI projects:

Lesson 1: Do not set too ambitious goals in the beginning

"Do not make the mistake of achieving hard results, such as direct financial gain, with AI projects," said Andrews. In general, it's best to start small-scale AI projects and focus on "soft" results like process improvements, customer satisfaction, or financial overview.

According to Andrews, early AI projects at best provide experiences that would help with later, larger implementations.

In some organizations, however, a financial goal is a prerequisite for starting a project. In this situation, the goal should be set as low as possible, Andrews said. “Think of targets in the five-digit range, understand what you want to achieve on a small scale, and only then you will see greater benefits.”

Lesson 2: Focus on giving more responsibility to people rather than replacing them

Great technological advances were often associated with a reduction in the number of employees. While lowering executive salaries is attractive, they are calling for resistance from those whose jobs seem to be at risk. As a result, companies can miss real opportunities, or use AI ineffectively. "We tell our customers that the greatest benefits of AI in the near future will be the ability to empower employees to engage in higher quality work," Andrews added.

Gartner predicts that by 2020, twenty percent of companies will deploy their employees to monitor and control neural networks.

"Forget about the idea of ​​huge teams of infinitely duplicable intelligent AI agents capable of performing tasks just like humans," says Andrews. It will be much more productive to motivate employees without leadership responsibility. Get them excited about how AI-driven decisions improve their daily work.

Lesson 3: Plan the knowledge transfer

Discussions with Gartner customers showed that most companies are not well prepared to implement AI. In particular, they lack internal skills in data science and they want to rely heavily on external providers to fill the gap. Fifty-three percent of the companies in the CIO survey rated their own ability to use internal data as minimal or very limited.

Gartner predicts that by 2022, eighty-five % of AI projects will yield erroneous results because the data, algorithms or teams are not optimal.

"Data is the fuel for AI, so companies need to prepare to store and manage even more data for AI initiatives," said Jim Hare, Research Vice President at Gartner. "The fact that you usually have to rely on external suppliers for these skills is not an ideal solution in the long run, so make sure that early AI projects support the transfer of knowledge from outside experts to your employees and build the internal capabilities of your organization, before you move on to major projects."

Lesson 4: Choose transparent AI solutions

In AI projects, software or systems are often used by external service providers. It is important to have insight into the way decisions are made and to integrate this into service agreements. "Whether an AI system delivers the right answer is not the only problem," says Andrews. "Leaders need to understand why it is necessary to have insights into the reasoning of AI systems."

Although it is not always possible to have all the details of an advanced analytical model, such as a deep learning neural network, it is important to offer at least a visualization of the possible options. In situations where decisions need to be verifiable, it may even be required to have new laws in Hong Kong to ensure that kind of transparency.