The manufacturing industry is not a uniform sector. There is the production floor, where physical work, machine operation and quality control come together. There is planning, which aligns orders, materials and capacity. There is administration around purchasing, inventory and invoicing. And there is customer contact, from quote request to delivery agreement. In this type of business, the largest part of office hours goes into planning, communicating and administering around the production process, not into the physical work itself. That physical work often still requires human hands, even though machine operation has long since stopped being fully manual.
What makes this sector special is the combination of many separate systems. ERP for planning, a separate package for inventory, a CRM for customer contact, and on the floor sometimes still Excel lists or paper work orders alongside digital systems. That fragmentation strongly determines what AI can and cannot take over well: a task that runs within one system is easier to automate than a task that travels through four systems and two departments. In addition, regulation plays a role, think of certification, traceability of materials and safety requirements. Where an error has consequences for safety or compliance, the bar for full automation is higher.
Within the manufacturing industry there are tasks that are largely rule-based and require little judgment. Think of sending standard confirmations after order placement, keeping track of simple inventory changes based on fixed rules, or generating standard documentation such as packing slips. Communication that depends little on the specific situation also lends itself to this. For instance, whether AI can take over sending welcome messages to new customers can often be done entirely by AI, because the content is largely fixed and the consequences of a small mistake are limited. The same often applies to sending reminders about appointments or deliveries, where it is also relevant whether AI can take over sending appointment reminders fits into that process. The common factor is: fixed content, low stakes in case of an error, and a clear moment at which the task starts.
The largest part of the work in the manufacturing industry falls into this middle category. Think of drawing up production schedules, where AI can make a proposal based on orders and capacity, but a planner must assess exceptions, rush orders and machine breakdowns. Think also of issuing work orders to the floor: a system can prepare and route the order, but whether this can be done without checking depends on how predictable the process is. The page about whether AI can take over issuing work orders to the production floor describes exactly where that boundary lies. Quality control based on sensor data or image recognition also often falls into this category: AI flags deviations, a human assesses whether that justifies a rejection and why. For these tasks, it is not so much the complexity of the action that counts, but the necessity of a judgment with a reason attached, something that requires oversight with approval or rejection rather than full automation.
Some tasks remain human work for now, and that is not a matter of preference but of the nature of the work. Physical actions on the floor, fine-tuning machines based on experience, and negotiating with suppliers about price and delivery time require judgment, negotiating room and physical presence that AI cannot provide. Complex complaint handling, where a customer reports a technical problem that does not fit a standard category, also requires someone who can ask follow-up questions and take responsibility for a solution. That is a different kind of work than the comparable customer contact tasks you see in what can AI take over in business services, where less physical context is involved.
The division between these three categories differs strongly per company, and even per production line within a company. A factory with one integrated ERP system and standardized products can place more tasks in categories 1 and 2 than a company that produces to order with varying specifications per order. The degree of regulation also plays a role: sectors with strict traceability requirements, such as food or pharmaceutical production, keep more tasks in categories 2 and 3 than sectors with fewer obligations. This pattern of fragmentation and regulation, incidentally, is also seen in adjacent sectors, as can be read in what can AI take over in the transport sector, where planning and customer communication likewise play a major role.
It is also important that this breakdown says nothing about personnel decisions. Whether and how an organization puts the freed-up hours to different use is a choice that lies with the employer, and one for which, when it comes to dismissal or reorganization, its own legal requirements apply. This page only describes which tasks are suited for takeover by AI and which are not, not what should be done with that.
To see how these three categories relate to your own tasks, you can fill in the free quickscan from ftetoai: twelve questions, no account needed, resulting in an indication of what portion of the hours in your profile can be taken over by AI today. The full work scan, which goes deeper into individual tasks and processes, is still under construction and is not currently being offered. The quickscan does provide a first, non-binding picture of where the room lies in your situation.
Vraag maar. Ik ken de kennisbank van deze site; wat ik niet weet, zeg ik erbij.
Answers come from this site’s knowledge base. Not tailored advice, and not a scan of your company.