AI is already disrupting the workforce. This will be felt most immediately through job losses, as systems start doing tasks previously assigned to people.
However, the level of automation involved in AI means, at least in the short term, there are a number of places where human-AI collaboration can work well. Those applications range from robotics in manufacturing and logistics to assistance in customer support.
Robotics and humans in manufacturing
Many manufacturing lines have moved to full automation, and robotics has played a key part. Back in 2016, a New York Times article pointed to Lawrence Katz, an economics professor at Harvard, who said automation would be a far larger driver of job loss than offshoring was. Nothing has changed.
However, there are some areas robotics can't address yet. In fact, a report by A.T. Kearney and Drishti stated that 72% of factory tasks are performed by humans. An interested follow-up apportioned the value of that work at 71%, showing there still is real value in human work.
Robotics can accomplish basic assembly line tasks very well, without the fatigue and errors of humans chained to the tasks. However, there are many qualitative and process tasks that are beyond the current technology skill level of robots. Basic quality assurance is easy, for instance. But vision still has a way to go for inspection, and the human eye can still better see other aspects of quality.
In addition, assembly of specialty items and cramped circumstances still require human dexterity. In a curious turnaround, Toyota's focus on human-AI collaboration means workers have returned to forging crankshafts. Their skills have resulted in material waste reduction of 10% and shortened the production line by 96%.
The clear fact that both robots and people have a place on the manufacturing floor means a lot of the work being done to embrace robotics today isn't directed at replacing people, but on how robots can safely work alongside humans. That is one area where networking and AI vision are helping.
Networking advances have enabled advanced deep learning systems to assist in the identification of people and robots, check their positions and have the mechanisms quickly adapt to the position of people.
Machine learning and humans in customer support
Customer support is another area where AI has been making inroads. While AI now serves a central role in the initial call system, as many studies have shown, people can still tell when they're talking to an AI and prefer talking to people. But that doesn't mean efforts to automate customer support are destined to fail.
Back in the dawn of AI time -- the mid-1980s -- I attended a graduate seminar where American Express was working to use expert systems to quickly analyze customer accounts so human call center employees could rapidly approve or reject large purchases for a card that had no set limit. Today, AI systems are similarly helping call center workers rapidly understand customers' issues and conditions, and to then suggest appropriate solutions. The person-to-person interaction still exists, while the ability to rapidly come to a resolution is improved.
There's another intriguing example of how AI can help. Got It is an AI company that focuses on helping customers effectively use features in Microsoft Excel. It has taken the concept of some freelance contract sites and wedded that to artificial intelligence. The end user, however, doesn't care about that; she only cares about the solution.
The Got It system begins with a chatbot, which self-identifies as such. It asks questions to identify the problem. As it gets the information, it uses machine learning to analyze the problem, compare it to the list of skills held by the contractors in the system and quickly link the client to the specialist. The company said its product, Excelchat, is able to link the customer with a live support person inside of 30 seconds 95% of the time.
The company is now extending its skill set to the wider business intelligence (BI) arena. Linked to both source data and experts, the Got It platform interfaces with the customer via natural language processing (NLP) to understand the problem. A base machine learning system starts with standard knowledge of the system, links the client to quick answers, communicates with experts to find answers to what's not already in the system, and then enhances the core AI system to learn from the resolved problem. Depending on the issue, the expert can be within the company or may be an external resource. Either way, the AI is working with experts to more quickly provide answers for harried end users.
"Artificial intelligence has great potential for working with human experts to enhance knowledge," said Peter Relan, chairman and CEO at Got It, based in Burlingame, Calif. "We are interested in how machine learning and NLP can integrate with knowledge experts in order to provide real-time solutions."
Got It's initial foray into the BI world isn't starting small. The first version of its product TransferML works with Salesforce and Google Analytics.
From these two very different scenarios, it's clear that human-AI collaboration can help people be more efficient. Yes, there are many jobs that will be replaced by AI. However, in many others, such as in manufacturing and customer support, AI will enhance the capabilities of humans. That doesn't always mean higher productivity with fewer people. In these scenarios, it can mean better service with the same number or more personnel.
What's clear is AI will change how we work. What's also clear is, in some ways, that will be a good change.