Illusion of Control
At a large metallurgical plant, a quality control department employee performed the same task for many years: counting and sorting metal parts coming off the conveyor after painting. He had experience, a good eye, and deserved authority with management.
However, the parts had neither experience nor authority. Hundreds of items were nearly indistinguishable from one another — the difference was literally a few millimeters, and there was no visible marking on them. As a result, due to the inability to accurately identify the product, the finished goods could sit in the removal area for up to six hours while employees figured out what the item was. The same problem existed in warehouses, receiving, and shipping across the country.
Inaccuracy in inventory accounting and control is not just a minor technical 'mismatch' or an inevitable cost of the warehouse. It is a stream of direct losses, wasted time and productivity, customer risks, and finally, strategic distortion of management decisions — something that can 'sink' even the most modern enterprise.
Three-Level Damage Architecture
Let's honestly break down what 'the simple human eye' and a pen in the accounting journal actually cost a business. Here, we can identify three levels of damage that accumulate like a snowball.
First Level: Direct Financial Losses
The most obvious layer: write-offs of shortages and surpluses, spoilage of goods due to 'lost' inventory. According to surveys, an average warehouse with up to 50 employees loses up to 8 million rubles a year just due to violations of labor discipline. And losses from manual inventory errors can reach up to 3% of turnover. Even 1-2% discrepancies in a turnover warehouse translate into hundreds of thousands of rubles in annual losses.
When an employee at the warehouse or factory receiving makes a mistake, money literally falls out of the accounting, and no one can find it.
Second Level: Internal Operational Costs
But the cost of a mistake is not just the amount of shortages. Real losses are almost always greater and consist of several components. When the data in accounting does not match reality, employees spend hours on 'investigations', manual checks, and unscheduled recounts. This time does not create value but is paid for by the business.
This is how delays in shipments and breakdowns in component supplies occurred at auto plants, and the same happened in food industry enterprises, where the warehouse did not have the goods that were listed in the documents.
Third Level: Strategic Risks and Incorrect Management Decisions
The most unpleasant and dangerous effect. When a company makes decisions based on distorted data, it either purchases excess items because 'the system shows a shortage' or fails to purchase necessary items because 'the system shows availability'. The digital reality ceases to reflect the physical one — and the business is heading for a cliff without realizing it.
Automation as a Magnetic Resonance Tomograph of Losses
Digital accounting using 'machine vision' or 'optical sorting' transforms the art of counting from a routine task into an exact science. It not only reveals losses but also immediately places them under strict financial accounting. Let's consider key breakthrough cases that have occurred in the last two years across several industries, clearly illustrating the value of transitioning from manual labor to automation.
Metallurgy: From Roulette to Deep Analysis
As in any heavy industry, the price of each mistake is high. For example, at the Pervomaisky Pipe Plant (PNTZ, TMK) in 2026, a machine vision system was implemented for continuous monitoring of the parameters of hot-deformed pipes. High-precision cameras, protected from extreme temperatures and dust, operate in real-time. The effect for customers is a guarantee of geometric size accuracy, while for the enterprise itself, it increases the metal utilization coefficient through prompt adjustment of equipment settings.
At the Magnitogorsk Metallurgical Plant (MMK), the accounting problem was solved differently. Every day, shop and warehouse workers manually counted pipes, with quantities in bundles reaching up to 600 pieces. This regularly led to discrepancies and sending consumers either insufficient or excessive quantities. Now, a neural network in conjunction with a web application on industrial tablets is used for this process. Labor productivity in the area has increased significantly, and the accuracy of accounting using specialized software has exceeded 99.6% (data as of 2025). The system recognizes pipes of various diameters and transmits information to databases in real-time with 100% accuracy.
Metal Scrap: When AI Comes into Play
Accounting for scrap metal coming for recycling is one of the most complex and contentious areas. Old control methods often led to disputes with suppliers, even resulting in arbitration lawsuits.
At the Vyksa Plant of the United Metallurgical Company (OMK), a machine vision-based service for quality control of steel scrap was launched in 2025. It works like this: when a wagon or truck with raw materials enters the casting and rolling complex, several neural network models analyze the video stream of unloading layer by layer and assess compliance with the stated GOST standards. One of the models can even stop unloading if explosive items — cylinders, pressure vessels, and similar — are detected in the scrap. Later, the service generates a detailed report for each vehicle. The result is the resolution of contentious issues with unscrupulous suppliers and accelerated acceptance of valuable raw materials.
Cases with Exemplary Efficiency
Sometimes automation integrates so deeply into the production cycle that the economic result can be measured almost in terms of money 'before' and 'after'. Two illustrative examples stand out here.
- GC 'Dobroflot'. An AI-based solution for accounting production was implemented on one of the canned food production lines. As a result, the speed of data entry about output increased sixfold, and the time for calculating employee salaries per shift decreased twentyfold (from 1 hour to 3 minutes). The annual economic effect on one line from replacing manual accounting and control of can seams amounted to 2 million rubles. The number of defective cans decreased by 17 times.
- AO 'APO Aurora'. At the moment of shipment, up to 10,000 50-kilogram bags of sugar are fed onto the conveyor belt daily. Manual counting of such a quantity of cargo provoked financial costs in the 'hundreds of thousands of rubles'. The video analytics system based on Vmx Qualex technology paid off in just six months, and now the counting of any bag in the flow is performed by the neural network with an accuracy of 99.99% (data from 2023, but the system is still operational).
What an Ideal System Looks Like: Accuracy, Speed, Awareness
Despite industrial specifics, modern optical accounting systems in production usually share common features.
- Extreme Accuracy. Regularly mentioned industry metrics range from 99.6% (MMK for pipes) in 2025 to 99.99% (in logistics for bags).
- Processing Speed and 24/7 Monitoring. Cameras and neural networks operate 24/7, capturing every object on the conveyor, preventing downtime and shipment failures, eliminating the risk of 'human' slowdowns and fatigue.
- Transparency and Objectivity. The system cannot make mistakes, get tired, or 'forget', and is also not subject to collusion. All data is stored in a common database, excluding information distortion. This allows for the recording of actual labor costs and provides an objective basis for management decisions.
Bottleneck: Human Factor and Illusion of Costliness
The most surprising thing is that systems do not necessarily cost billions. For example, the development by MTUCI scientists, ready for industrial use, can recognize types of pipes in photographs taken with an ordinary smartphone, making the technology accessible for small and medium businesses, while an advanced computer vision system on the conveyor can pay off within the first months after installation.
Nevertheless, the key factor resisting implementation remains not the price, but ordinary human skepticism, fatigue from change, and, paradoxically, the illusion of control. Shop and warehouse managers believe that their employees count accurately enough and that all figures add up. However, as practice shows, even with regular inventories, the error rate in simple visual control easily exceeds 5-7% of turnover and sometimes reaches a quarter of the total volume. At many enterprises, they are simply written off as 'natural technological losses', which seem inevitable — and disappear into thin air.
The true picture of losses only becomes clear after implementing an objective accounting and control system.
From Losses to Profit
Recent data speaks for itself: for enterprises with manual accounting, up to 95% of errors can be eliminated by transitioning to an automated system. Cameras and neural networks displace the human factor from an area where it does not belong — from a monotonous, resource-intensive, but critically important process for the financial health of the enterprise.
Losses and leaks from unaccounted inventory amount to millions and billions of rubles across the industry. When an enterprise is ready to acknowledge that the 'human eye' is insufficient, optical sorting and accounting will help quickly catch up with competitors who are already counting every percentage of profit using 'digital vision'.