Why AI isn't solving manufacturing's productivity problem

For the best part of two decades, manufacturers have been buying productivity. They have modernized enterprise resource planning (ERP), automated production lines, digitized supply chains, layered in predictive maintenance and optimized more processes than most people can name.

AI is simply the latest wave of that same instinct, and the appetite is real. Deloitte reports that 80% of manufacturing executives plan to put at least a fifth of their improvement budgets into smart manufacturing, and Gartner finds that 76% of CEOs believe AI is the technology most likely to disrupt their industry over the next three years.

Those numbers don't tell you AI works. They tell you the money and the expectations have already moved, which makes the next question the important one: what actually gets in the way of a return?

Because for all of it, work still slows down. Not for want of data, and not for want of intelligence, but in the gaps between systems, teams, and processes, where a good decision has to be turned into a completed piece of work.

Most of the decisions already exist. Production plans exist, demand forecasts are available, inventory positions are visible, and pricing models are more sophisticated than they have ever been. The gap is not in knowing what to do. It’s in getting it done, and once you look at the problem that way, it starts to show up everywhere. Manufacturers don’t have an AI problem, but an execution problem.

Every manufacturer has lived this day

Take a fairly ordinary Tuesday morning. A supplier moves a delivery date. A key customer asks for expedited fulfillment. Material costs shift, inventory drops below target, a pricing exception needs sign-off, and an ERP change is queued before the next release.

None of this is unusual, and none of it is caused by missing information; in most businesses, systems flag each of these events almost the moment they happen. The difficulty starts after the alert. Who needs to know? Which system gets updated? Does procurement need to find an alternative supplier, does production need rescheduling, does the price change, should customer service get ahead of the delay, does finance need to revise the forecast?

None of those questions are hard to answer on their own. What is hard is answering all of them at once, across functions, before the day runs away from you.

Picture that same supplier slip in a business where the coordination actually holds together. The delivery change lands, and instead of triggering a dozen emails and a scramble, it flows straight through. Procurement is prompted with a ranked set of alternative suppliers. The affected production run is automatically resequenced against the new dates. The customer team gets a drafted heads-up before the customer notices. Finance sees the revised cost feed into the forecast, and a human signs off the one decision, the pricing exception, that genuinely needs judgement.

Same alert, same information. The difference is that the work between the alert and the outcome doesn't fall on the floor, and that is where most of the value is still sitting untouched.

None of these problems are new. What's changed is that AI makes the quality of ‌decisions better—but it doesn't automatically improve the way work moves through the organization. Until organizations address that execution gap, better intelligence alone won't translate into better business outcomes.

Leading manufacturers aren't simply adopting more AI

It’s also what separates the best operators from everyone else, and the gap is widening. LNS Research found that the world's most productive industrial companies delivered 13.9% greater productivity growth than their peers over the past five years, and improved operating margins by 18.5 percentage points.

The figures don't, on their own, prove why, but the pattern is hard to miss once you spend time inside these businesses. They don't optimize pricing, procurement, inventory, and production in separate boxes; they coordinate those decisions, because in practice the decisions are never really separate.

Commercial choices shape supply chain choices, supply chain choices shape customer outcomes, and information has to move continuously between systems, people, and processes for any of it to hold together. What they've built is faster execution, optimized across the whole business rather than in one corner.

Which is why I think the industry's instinct to reach straight for the model is slightly misplaced. The conversation almost always drifts towards which models to use, how to govern them, where to deploy them, all fair questions, but rarely the ones actually holding organizations back.

Gartner's own research hints at where the discomfort really lies: while 76% of CEOs expect AI to disrupt their industry, only 44% believe their CIO is AI-savvy enough to lead the transformation. Read that honestly, and it is less a verdict on CIOs than an admission that even the people meant to lead this don't yet feel ready, not because the technology is missing, but because nobody is sure how to make it stick to the day-to-day work that actually moves the business.

AI without execution creates a bottleneck

Another model will not resolve a supplier issue, approve a pricing exception, update the ERP, notify the customer and reschedule production. Something still has to coordinate all of that, and if nothing does, the intelligence just piles up.

Maxim Ioffe, Senior Director, Intelligent Automation at Wesco, framed the risk as well as I have heard it: "Without thoughtful orchestration, the productivity gains achieved by AI might simply shift the bottleneck to the next step of the process and have minimal impact on the overall speed of execution."

That is exactly what is starting to happen. AI produces better recommendations, but a recommendation on its own creates no value; it only becomes worth something when it is connected to the work. That connection is precisely what most manufacturers are missing.

Look at the technology estate and the shape of the problem becomes obvious. Manufacturers already have systems that store information, such as ERP, manufacturing execution system (MES), planning, and CRM, and they are investing quickly in systems that reason, from AI assistants and predictive models to generative AI and digital agents. What sits between the two, almost always, is a void.

There is no layer that coordinates the work itself, connecting recommendations, enterprise applications, automation, documents and people into a single, governed process. That missing layer is, increasingly, where the advantage lives.

It is the least glamorous part of the stack and, for now, the most valuable: the difference between the two Tuesdays described earlier.

The missing layer in manufacturing AI

It’s worth being honest that this is harder than it sounds. Coordinating work across functions is exactly the kind of thing organizations have been promising themselves for years and rarely delivered.

The layer only matters if it can actually reach across systems, hold a process open over days rather than seconds, and keep a person in the loop where judgment is needed. That is a high bar, but it is the right one, and a far more useful thing to be arguing about than which model to buy.

That is increasingly where competitive advantage is emerging.

As Raghu Malpani, Chief Product and Technology Officer, UiPath, recently said on our podcast, “Orchestration is the difference between a company owning a fleet of trucks versus owning a logistics company. The trucks do the work, but the logistics company decides what the trucks do, where they go, how they handle a flat tire or a driver falling sick or a flood happening in a route. Orchestration is the logistics company—and the workhorses are the tasks.”

What manufacturing leaders should focus on next

Manufacturers don't have a shortage of data or a shortage of AI models. Increasingly, they won't have a shortage of recommendations either. The organizations that pull ahead will be the ones that consistently turn those recommendations into action.

That means coordinating work across commercial teams, supply chains, finance, customer service, and operations instead of allowing each function to optimize in isolation. It means connecting intelligence to execution so decisions move through the business as quickly as the information that created them.

That's where the next wave of productivity gains will come from. Not from making AI smarter. From making the business better at acting on what AI already knows.

Continue reading

If you are thinking about how to move past AI pilots and connect intelligence to real business outcomes, we go deeper in our new e-book, The Missing Layer in Manufacturing: how manufacturers are connecting AI, enterprise systems, automation, and people to execute work across the business.