CT defense giants urge measured approach to AI implementation

Connecticut’s manufacturing suppliers should focus on collecting reliable data and protecting sensitive information as they prepare to adopt artificial intelligence, according to executives at three of the state’s largest defense manufacturers. A panel discussion at the Connecticut Business & Industry Association’s Manufacturing Summit brought together AI experts from Sikorsky, Pratt & Whitney and Electric Boat, to discuss the future of the technology in manufacturing. All three offered warnings about the speed at which the technology is changing and the need to remain flexible about integration into longstanding processes. Each company is adopting AI through a mix of purchasing off-the-shelf platforms, partnering with AI specialists and building custom software applications. Gary Cooper, director of combat and weapons systems engineering at Electric Boat, said no matter the source of the technology, effective implementation requires the buy-in of key staff in the organization. “The real power of agents and assistants really comes from subject-matter experts in your organization that know your business sitting down with the tool and building capabilities in a very, very rapid fashion,” he said. He also cautioned against partnering with AI startups without first determining what expertise they offer and ensuring they are not simply experimenting with company data. Pratt & Whitney has also pursued acquisitions to expand its AI capabilities. It recently expanded its engine inspection capabilities by acquiring Amsterdam-based Aiir Innovations, which had developed AI-assisted software. All three are leading the way for their supply chain companies in AI implementation. Pratt & Whitney recently polled 1,500 supply chain companies and found that around half are actively exploring or implementing AI. The three panelists agreed that most companies should begin by capturing, organizing and validating data. That requires building digital design and data-sharing capabilities that reduce reliance on paper blueprints. “Just thinking about data being accessible across every phase of the business,” said Peter Bradley, who leads digital tools and data science at Pratt & Whitney. “There's just no getting around the fact that some of this is really about the grind. All that flashy AI stuff doesn't really happen without a lot of that foundational work.” Paul Link, program manager for AI and data at Sikorsky, echoed that point. He said suppliers should first identify problems in their production processes. “Starting with understanding your pain points — where you have challenges, bottlenecks, inefficiencies — and then evaluating AI as a potential solution is critical,” he said. Cooper said Electric Boat has been hiring data scientists in its efforts to benefit from AI. “We have 125 years worth of data we've built our company around and we need to structure it better,” he said. A major part of that effort is also data security for companies that handle sensitive government contracts. “We personally have to make sure that our data doesn't train the world how to build submarines,” Cooper said. Workforce training and strategic hiring are also key to adopting AI successfully. Bradley said some Pratt employees with expertise in areas such as aerodynamics or materials are returning to school to earn an AI degree. “That's allowing them to innovate in new ways,” he said. Cooper said some of Electric Boat’s waterfront workers, including mechanics, welders and machinists, may not interact with AI directly, but they can still benefit from improvements behind the scenes. “What I see right now is us helping the mechanics with AI by having these simpler, easier-to-understand work orders in front of them and having the material where they need it, when they need it.” he said. Still elusive is a concrete way to measure the overall return on investment for implementing AI, although Pratt’s Bradley said in some individual applications, the advantage is obvious. For instance, he said Pratt now uses AI to help take complex measurements of small parts during inspections. “We're able to do things in seconds that used to take minutes,” he said. “The other thing is the way they were doing it before was really painful for the operator, really tedious.” Bradley said that despite all the fundamental changes AI may bring, companies should keep the technology in perspective and tune out the hype. “It's really important to understand that this is not magic, it's not even intelligence,” he said. “It's really math and software.”