Manufacturers are pouring money into AI, robotics and factory automation, expecting the technology to help them make more products, cut costs and stay ahead of competitors. Buying the equipment is often the easy part. Getting it to work properly inside an actual business is usually where the trouble starts.
A robot might perform well in a demonstration. An AI system might spot defects with impressive accuracy in a test environment. But those results do not carry automatically into production. If the robot cannot work alongside older machinery, or the AI cannot communicate with equipment that was never designed to connect with it, that is where costs businesses did not plan for start showing up.
Manufacturers have to think beyond the equipment or software they are buying. They need to consider how everything fits together, whether existing production lines can accommodate the new technology, and what happens when something goes wrong. Employees need training, production schedules may need to change, and engineers often find problems that nobody spotted during the planning stage. For a business expecting quick savings, those complications can be expensive.
The price tag rarely tells the whole story
When a manufacturer approves a new technology investment, the purchase price is usually the number everyone looks at first. The less obvious expenses tend to come later.
Also read: Jim Cramer sends a strong signal to Nvidia stock investors
A company bringing robots onto an existing production line may need to modify machinery, install safety equipment, develop new controls and make sure components from different suppliers can communicate. Specialized tooling, employee training and testing all add to the bill before the system is ready for production.
Steve Mai, CEO of Eclipse Automation, which designs and integrates digital manufacturing solutions, told TheStreet he sees those hidden costs play out regularly.
“Almost nobody loses money on the robot itself,” Mai said. “They lose it in the months between the robot arriving and the line actually running at speed, because nobody budgeted for the work of making it fit into everything else on the floor.”
McKinsey has warned that poorly planned automation programs can become expensive, take longer than expected and fail to deliver their intended results. The real question is not simply how much a piece of technology costs. It is how much the company must spend to make it useful, and whether the eventual savings justify the full bill.
AI is running into the realities of the factory floor
The growing interest in AI is adding another layer to the challenge. Manufacturers are exploring ways to use the technology to inspect products, monitor equipment and improve production, but moving from a promising pilot to reliable everyday use has proven difficult for many businesses.
A McKinsey survey of more than 100 manufacturing chief operating officers at companies with at least $1 billion in revenue found that roughly two-thirds said their companies’ AI implementation remained at the exploration or targeted-implementation stage. Only 2% said AI was fully embedded across all operations. Another 46% reported limitations in their data or technology systems.
Take an AI system designed to spot defects on a production line. The software might be excellent at recognizing problems, but the manufacturer still needs the right cameras, lighting, machine vision equipment and controls. The system must also work consistently at production speed and handle changing conditions. If it cannot communicate with existing machinery, its detection capabilities may not deliver much value.
“A pilot proves the AI can see a defect,” Mai said. “Production asks a harder question: can it see that defect on every part, on every shift, across multiple plants at full line speed, and then tell the machine what to do about it? That’s an engineering problem, not a software problem, and it’s where most projects stall. To get past it, you have to update the model so it understands the action to take the next time the same issue is presented, then give it the freedom to act within those constraints.”
When a technology project starts eating into profits
The cost of a troubled implementation goes beyond an unexpected engineering bill. If a factory upgrade takes longer than planned, the company may have to keep older equipment running or delay a production change. If a new automated line keeps stopping, employees end up doing work the system was supposed to handle.
There is also a risk in automating a process before fixing what is wrong with it. If a company’s workflow is inefficient, adding automation may simply make it easier to repeat the same mistakes at greater speed. The business ends up with a more complicated system but not a better result.
Those problems can erode the business case quickly. A project that looks attractive on paper may take far longer to pay for itself if the equipment is not operating as intended. Management that expected productivity gains to arrive within a year may find itself waiting considerably longer.
Deloitte’s 2025 smart manufacturing survey found that 92% of respondents believed smart manufacturing would be a main driver of competitiveness over the following three years. Manufacturers also reported productivity and output improvements from their initiatives, though results varied. Higher spending does not automatically translate into higher productivity.
Why getting everything to work together matters
One way manufacturers can reduce the risk is to pay more attention to the engineering that connects software, machinery and other equipment. Robotics, machine vision, controls and AI are often discussed as separate technologies, but in a factory they have to function as one system.
When several suppliers are involved, manufacturers need to establish who will handle compatibility problems, who is responsible when the full system does not meet expectations, and who will support the operation after installation. Otherwise, a problem can turn into a drawn-out dispute over which supplier is accountable. Having a clear answer to that question before work begins matters more than most procurement teams expect.
More Tech:
- Apple CEO signals surprising change for iPhone users
- Comcast sends wake-up call to employees as customers cut services
- AI agents’ growing role in payments raises an awkward question
“Every vendor can show you how their piece works,” Mai said. “The question to ask before you sign is who owns the result when all the pieces are running together. If the answer is nobody, you’ve just found the most expensive line item in your project.”
Manufacturers also need to think about what happens after a system goes live. Equipment requires maintenance, software needs updates, and production requirements can change over time. A solution that works well on day one may need adjustments as the business evolves.
The decision companies need to get right
Businesses are unlikely to slow their investment in factory automation as they look for ways to improve output, control costs and compete. But the manufacturers that get the most from these investments may not be the ones with the newest equipment or the highest AI spending.
That means defining the problem first, calculating the full cost of implementation and confirming the new system can work with existing operations. Training, maintenance and ongoing support should be built into the plan rather than discovered as extra costs after the budget has already been approved.
For investors, this matters because technology spending does not automatically translate into better margins. The return depends on whether the business can deploy the technology successfully and generate savings or additional output that justify the expense.
The expensive mistake is not always choosing the wrong technology. Sometimes it is underestimating everything that needs to happen after the purchase order is signed.
Related: Mark Cuban warns investors to brace for an eventual market turn