When you think of smart automation, what do you think about? Lights-out factories? Completely autonomous production lines? Robots running the plant? Or do you think of something more data-centric and predictive? Perhaps even something involving AI, AR or MR? Turns out, none of those are wrong. There’s a lot of moving parts that go into the idea of smart automation…but we’re getting closer to unlocking its potential.
Of course, the technology has to exist before implementation. Much of the time, individual intelligence and automation technologies already exist. The key is bringing them together. This was a focus of an Automate Forward panel discussion.
“[Smart automation] is no longer viewed as the crazy scientists in the lab,” said Tom Panzarella, senior director of perception at Seegrid. “[It’s] actually solving a business problem, actually quantifying it for the business and then treating the technology not as the end, but rather the tool.”
With a more level-headed mindset on automation potential and data, companies are fine-tuning their infrastructures in preparation for smart automation. This is expected to create a “tipping point” in which it becomes easier to implement complex statistical analysis and AI into production lines.
“The algorithms themselves in many cases aren’t new,” said John Lizzi, executive leader of robotics and autonomous systems at GE Research. “But it’s really all the infrastructure, the tooling and the frameworks that have made this easier.”
The automation journey isn’t a straight path, especially with safety critical applications that have to adhere to safety and governmental regulations. Rashmi Misra, head of AI, mixed reality and silicon business development at Microsoft, discussed how a lack of automation guidelines and standards can lead to deficiencies in the application because engineers are literally left to their own devices.
“We’re all in an ecosystem having to work together under those same conditions,” she said. While use cases arise, there are variants of cases that need to address individual automation business models and goals. She noted that some established use cases, or toolkits, can be adapted for another business model.
Toolkits can serve as a reference guide for applications along the automation journey and can be valuable to small- to mid-size companies that don’t have a large research division. Rishi Vaish, CTO and VP of IBM AI Applications, explained two different levels of investment IBM makes in order to make its automation technology consumable.
“The first one is in the tooling,” he said. “One level of investment is continuously making those tools.” This includes data, the model, keeping the model running in production and enabling the model to measure bias in the system.
“The second level of investment is when we actually build an application,” he explained. “For most companies that just want to get going, some higher level of abstraction is a much faster way to get their AI journey kicked in.”
Large and small manufacturers alike share a similar hardship, though: effectively integrating the new tech with the old tech.
“For us, that’s where the challenge is at,” said Jorge Ramirez, global director of execution automation and chief manufacturing cybersecurity officer at General Motors. “We are limited by capital. The easy solution would be to take out all the old and put in the new with all the new smarts that just play harmoniously.”
We all know that’s not the reality, though.
The convergence of legacy with smart technology is one of the biggest challenges in smart automation, but it’s occurring more often, which will lend a hand to toolkit expansions, specified use cases, and eventually, to easier global adoption.
Lizzi advised companies to look at both the strengths and weaknesses of a system to determine where smart automation could live. It’s also important to look at the strengths and weaknesses of the smart technology too.