Every time a company tells me about their efforts to implement IoT projects, I always ask them about the challenges they faced along the way. Among the most consistent challenges (outside of connectivity) is getting employees to adopt the new technology and in the process, change their ways of working. It’s as though in their rush to adopt a technological solution to improve their business, managers forgot that those employees would be handling any information delivered by the newly connected machines and so they need to buy into the insights delivered. Or that employees might be miffed at being told how to do something by a computer. Or even that those employees might have input that could help make a computer better.
To counteract the efforts of Luddites who rebel at the hint of technology, and to avoid missed opportunities caused by not involving workers in the implementation of a digital transformation, researchers from Intel have figured out how to combine workers and AI in ways that make everyone happy.
Intel’s Irene Petrick, director of industrial innovation, and Faith McCreary, industrial IoT principal engineer, user experience architect, and researcher, have spent the last few years poring over data gathered from the company’s manufacturing customers. They’ve also been talking to Intel customers about their expectations, efforts, and hands-on experiences when it comes to making their factories more automated and intelligent. Intel — both as a manufacturing company and as a provider of baseline tech for the intelligent factory — has a huge stake in understanding the problems of digital transformation.
The first research report from these two women came out in April 2018 and focused on the roles of people inside intelligent factories. For their next research project, which should be made public before the end of this year, Petrick and McCreary looked at how workers will adopt and react to AI in manufacturing roles.
At Intel, where the manufacturing process is very regimented and conducted in literal clean rooms where employees wear bunny suits and are forbidden basics such as makeup for fear of contamination, AI and true intelligent factories where people aren’t necessary will benefit yields and the bottom line. But the researchers wanted to know what happens when people have to stay and work in an increasingly automated manufacturing environment.
McCreary says the two have a stack of responses 77 inches high from manufacturing leaders, executives, and factory workers. And as it turns out, much of the commentary in those responses as it relates to AI comes down to trust. The secret to getting people to accept and adopt AI in a factory, and maybe in an enterprise, environment isn’t rocket science. It’s common sense.
When it comes to working alongside machines and AI, workers want to understand how the algorithms work — or, at the very least, how those algorithms were trained and what problems management thinks the AI is trying to solve. “The challenges to deploying smart manufacturing aren’t technical,” says Petrick. “Almost universally, people want to know how the AI comes up with the algorithm.”
Also, when it comes time to implement AI, roll it out slowly. Let someone work alongside the machine or rely on the algorithm for a while before turning over more of the work handled by people. Then, when the process gets more automated, make sure someone oversees it. At that point, if everything is working as planned, turn over the entire process, and ideally have training in place so your employee can step up to a more complex role in the operation associated with ensuring the algorithms are still working as intended.
When the effort is successful Petrick says the survey respondents say it’s like getting a new teammate that they can rely on to improve operations.
There’s another thing to think about. When budgets are cut in relation to a digital transformation that combines AI and IoT, among the first things to go are change management efforts and worker training. But when you’re trying to get workers to follow advice from an AI, such training is critical.
Budget cuts and even a related lack of focus are also why IoT-related pilots may not scale widely across an organization. “Anytime you roll out a project where there’s a new way of doing things, you need training. If workers don’t actually do something different then all of that data and insights are worthless,” says Petrick.
The two have many more insights to unveil based on their research, including segments about AR adoption and why picking a use case may not be the best way to start an IoT project. We’ll hear more about them in future issues as they share more information. I can’t wait.