RCCN WeChat QrCode
Mobile WebSiteThe era of artificial intelligence to create business value has arrived. In fact, in a recent survey by Deloitte, 82% of early organizations using artificial intelligence said they had achieved economic returns from artificial intelligence investments.
Artificial intelligence and related technologies are improving existing products and creating new ones. These technologies are optimizing internal and external operations, helping organizations make better decisions, making employees more creative, engaging in higher value jobs, and bringing multiple benefits.
There is no doubt that 88% of companies plan to increase spending on cognitive technology in the coming year.
However, artificial intelligence is not a panacea for all business problems, and the use of artificial intelligence technology is no easy task. Here are the most important challenges companies must overcome before they can see the positive results of deploying artificial intelligence technology.
Data problem
The biggest obstacle to starting an AI project is data. Specifically, there is a lack of available and relevant data that is not inherently biased and does not violate privacy rights.
According to Deloitte's survey, 16% of IT executives ranked data issues as the biggest challenge associated with artificial intelligence, higher than any other issue, with 39% of respondents citing data in the top three worrisome aspects. .
Raj Minhas, head of the Artificial Intelligence Research Laboratory at the Palo Alto Research Center (PARC), says many companies will collect data as part of their day-to-day operations. "But these data may not be the right data."
Before launching an artificial intelligence program, companies must carefully study the data they have and look for areas of higher value.
“It's like looking for a lost key near a streetlight, not looking for where you lost your key,” he said. “We recommend that companies look back and see where they can get the most value, rather than looking at where they have the most data.”
Another problem is that there is no suitable amount of correct data.
“We work with many customers with large capital infrastructure, such as wind turbines and rail systems,” he said. “All of these devices are designed to be very reliable.” Therefore, when companies try to use machine language to predict failures before they occur, they find that 99.9% of the data collected from these devices comes from their normal operation.
"What you are concerned about is the abnormal behavior of the machine," Minhas said. "So, you have a lot of data, but these are the wrong data."
Business process challenges
How to integrate artificial intelligence technology into the company's functional departments is another obstacle and is listed as the second largest issue in Deloitte's survey.
Vivek Katyal, global head of analysis and data risk at Deloitte Risk and Financial Advisory, said: "Structural and cultural factors remain one of the key factors hindering the application of artificial intelligence and one of the biggest challenges. "People are still trying to understand the impact of artificial intelligence, what it can do, what it can't do. It's like a terrible robot breaking into an organization and messing things up."
He said that when artificial intelligence is embedded in a platform that people already use (such as ERP or CRM systems), this application will be easier. In fact, people may not even know that artificial intelligence technology has been used.
“But when we talk about artificial intelligence changing business processes, it fundamentally changes the way companies work and what they do, and it's an area where there are more difficult issues to solve,” he said.
Technical implementation challenges and skills shortages
The implementation of artificial intelligence technology has brought many technical challenges. Most organizations do not have enough artificial intelligence skills to skillfully deal with these challenges. In Deloitte's survey, 39% of the respondents ranked the technical issues as the top three. One of the challenges, while 31% of respondents ranked lack of skills as one of the three major challenges. In addition, 69% of respondents indicated that the shortage of AI skills is moderate, severe or extremely severe.
Svetlana Sicular, an analyst at Gartner, said: "The current situation is that most companies can't solve these problems themselves because they don't have these skills." A year ago, when she talked to business users who were just starting to research AI At the time, most companies thought they would build their own systems. By the end of the fall, this number has changed, and now about two-thirds of companies want to deploy AI by using embedded tools in smart enterprise applications. “The situation is changing very quickly,” she said.
It is one thing to make this technology work; it is another thing to make it work in the real business.
“Many companies are not ready to accept the fact that the output of machine learning technology is probabilistic,” Sicular said. “Some of the results are always incorrect. For them, this is completely unexpected. They need to design for exceptions and provide some means for feedback loops.”
Tool and development costs
For organizations that build AI systems from scratch, the cost of labor and technology can be high. This is especially true for organizations that are just getting started.
The Anderson Center for Autism in northern New York has 850 employees and its chief information officer, Greg? Gregg Paulk said: "When I first entered this company, we took this path."
He said that the establishment of a new artificial intelligence system is very expensive both in terms of capital and personnel. “We are a small non-profit organization. We don’t have these developers.” So, for the center, like many small organizations, this means having to hire an outsourcing company to do the job.
“In the past, we have been trying to do something similar, because of the relationship between cost and development time, we failed,” said Paulk.
Instead, the organization is applying AI tools to systems already in use by the company. For example, the Human Resources platform from Ultimate Software now supports artificial intelligence-driven tools that allow organizations to investigate employees, including asking open questions and using natural language processing and sentiment analysis to intelligently analyze responses. . The software also advises managers to take concrete actions to address employee issues, which has caused employee turnover to drop by more than a third in the past two years.
“In 2013, when they first started discussing artificial intelligence at the conference, I thought, ‘this technology will never be used,’” said Paulk. Now he is "surprised" by the capabilities of the technology and the organization is already using it through cloud-based systems.
"We certainly can't do it ourselves," he said.
The situation at the Anderson Autism Center is not a case. According to Deloitte, 59% of companies acquire AI technology through enterprise software vendors. For example, Salesforce Einstein is a built-in AI tool that helps sales representatives determine which potential customers are more likely to convert to actual buyers.
49% of companies use cloud-based AI. Many vendors and cloud providers offer off-the-shelf AI services, so organizations don't need to build their own infrastructure or train their own algorithms.
Both of these methods can reduce costs or shift costs from the IT department to each business unit. For cloud applications like Salesforce, there is less demand for physical infrastructure or internal support work or management because most of the work is handled by vendors.
