It’s no secret that across US industry, adoption rates of AI, and technology in general, differ. Consider service giant Amazon. Insiders now forecast the company to jettison 500,000 employees in the coming years because of AI, on top of tens of thousands already sent packing. 

Manufacturers have no such luxury – or inclination. People are gold. And if manufacturers lag behind their service counterparts in deploying AI, there’s often a good reason, namely the big difference between moving a product and making one. 

But with so much buzz attending to AI, it’s understandable if companies are nervous. With so many stories of cost savings and enhanced profitability floating about, it’s easy to worry about being left behind in the AI scramble. 

Art Casasa is a Houston-based Senior Engineering and Program Management Executive who helps organizations bridge the gap between traditional manufacturing and the future of intelligent, data-driven production. He told me manufacturers shouldn’t lose sleep over being left behind. Yet.

“It’s not a bad thing that manufacturing is not adopting at the same rate (compared to services),” Casasa said, “because if they were, they’d be failing. A recent study at MIT indicates that 95% of implementations of AI fail – completely fail. So you only see like five to 10% good implementations,” he says.

That’s not to say that more could be succeeding, with a different approach. “What’s happening right now is companies want to put a layer-like cake of AI on top of their systems. ‘Oh, let’s do AI here and there and there and here.’ And let’s grab all the ideas from the workforce on where they or we should put AI.’ However, they’re not looking at the value proposition,” Casasa says. “And that’s when they fail. And like I said, like 90%-plus are investing too much money in AI. They’re pouring money into AI like crazy. But it’s not proven. And the reason it’s not proven is like they want to do everything – and that’s not possible. It’s like any manufacturing process. It takes time to onboard new processes, including AI,” he says.

Casasa points to another barrier – people. “Often the most difficult part (in succeeding) is onboarding the people that need to use AI.” Call it another irony of the AI revolution. 

“Companies want to follow the trend, right, where AI replaces middle management, reduces top management, and companies then move into a state where there’s more self-administration, where there might be one person here or there, aligning the workforce,” he says. “But what I think will happen, is that the workforce will start doing their own administration. And people on the lines, technical people on the lines, like you mentioned – technicians, operators, all of these people – will remain. So anyone that’s working with their hands, if their work cannot be automated, which is many cases, maybe 80% – they’ll stay.”

Yet some employees, in all companies and across industries, are in the crosshairs according to Casasa. “I can build an (AI) application of checking invoices versus purchase orders and find problems in between before they hit the payment cycle. That’s something that a person might take four or five hours a week. Now it’ll be two minutes,” he says. “You’ll start seeing that and you’ll start seeing accounting being reduced because now you can do the full accounting of a company by an agent in AI, that will work better than a CPA. Those are facts. Accounting will be displaced, HR, to a point, that will be displaced, and all the processes within hiring and posting, all of those things can be done with AI,” Casasa says.

“I think what companies need to do, they need to start working on leadership at the lower levels, because I don’t see the middle management surviving in five to ten years.”

I speculated that one interpretation of Casasa’s analysis – one path forward for manufacturers –  is to dive into the operational parts of the business where AI can be leveraged now, like backoffice savings, then pour those resources into technology adoption in the production side of the business. Manufacturers still need to fully adopt pre-AI technology, after all, like IoT. 

Casasa went along, to a degree. “IoT, like you mentioned, that was 4.0 kind of thing, it should have been implemented to the maximum ability – right now. It’s actually not difficult to implement. The first stage is going to be heavy on cost, but it’ll pay for itself within a two year cycle, maybe even a one year. Implementing IoT, and then dumping all your data into the cloud, is a relatively easy process. It’s just depending on the amount of IoT devices you have, but it’s just scaling the cloud to whatever you have in the ground. It’s actually not difficult to do.”

If you’ve not found it easy, Art Casasa and others – including GHMA contributor Nadeem Azhur – can help. Azhur’s monthly blog is great reading, and Casasa will begin leaning in with a series on AI implementation, in the coming weeks.

For those manufacturers that have kept their powder dry, congrats. But there’s no time to be complacent. 

Bart Taylor is executive director of the Greater Houston Manufacturing Association and founder of Inside MFG. Reach him at [email protected].

Contact Art Casasa at [email protected].