Say, ERP system, when will the delivery arrive?

Change cycles for production processes and other company processes are becoming shorter and shorter. Conventional ERP software, however, is not exactly known for its adaptability or user-friendliness. Unstructured data is generally indigestible, leading to tedious data entry. Errors often go undetected and customer requirements fall by the wayside. But it is precisely these weaknesses of classic ERP systems that are the strengths of analytical and generative AI.

The obvious quick wins of intelligent ERP

Data capture: With the right prompts, generative AI can extract complex data structures from catalogs, descriptions and other continuous texts with increasing reliability and capture them directly in ERP systems.

User-friendliness: Programmers can now code many times more software with AI-driven code generators. Soon, such generators will generate individual input masks and other user interfaces on the fly that are tailored to the individual user and their use case. Gone will be the dictatorship of standardized input masks that are a little bit suitable for every use case, but not really suitable for any of them.

Flexibility: Today, company processes are coded and configured more or less manually in software. AI will not only suggest independent optimizations, but will also be able to implement completely new processes on demand.

Individualization: Henry Ford used to offer his cars in all colors as long as they were black. Increased individualistic customer requirements are generating ever greater demand for personalized products. Before the ERP era, it would have been impossible to cope with such increasing complexity in information management. Today, at worst, this requires a new table in the system or something similar.

Anomaly detection: Modern AI is particularly good at detecting subtle deviations in large data streams. In this way, complex errors are detected at an early stage. This applies in particular to automatic data acquisition and software, which are only made possible by generative AI. Even misuse, fraud and waste are easier to identify with analytical AI.

ERP systems are transaction costs that have become software

When Adam Smith discovered economies of scale in the markets of the 18th century, there were no large companies. Economies of scale are enormously effective, but not very intuitive. As a result, they became one of the most overrated concepts in business administration.

Large companies only emerged in the middle of the 19th century because it became possible to limit liability at that time, but AI did not yet exist. Since companies and other organizations are smarter than any individual employee, they are rightly referred to as an early form of superintelligence. This decoupling of decision-making and ethics enabled the heyday of its success in the 20th century.

As recently as 1930, Nobel Memorial Prize winner Ronald Coase discovered that economies of scale are attributable to transaction costs. The introduction of ERP systems represents the digitalization of transaction costs, which has reduced them enormously. The lower the transaction costs, the lower the economies of scale. Otherwise, the restructuring of the entire automotive industry would hardly have worked.

Smart ERP systems are turning the economy on its head

People often forget how revolutionary ERP systems were when they were first introduced. The automotive industry has been demonstrating this upheaval for decades. The introduction of the first ERP systems has transformed formerly huge car manufacturers. The once monolithic economic structure has disintegrated into many smaller suppliers. This does not fit in with the naïve view of economies of scale.

Anyone who has ever tried in vain to buy from a sister department at favorable market prices knows the reason for this. Vain full-cost accounting dominates within companies. But low market prices are so called because they are only possible in a competitive market environment.

Markets and later also companies are simply two opposing methods of organizing human work. Competition on markets does create a price advantage, but also an information disadvantage.

Anyone who has been in the market for used cars and wondered whether the information on the speedometer is correct, the spark plugs are really new and the bodywork is rust-free under the paint is familiar with the disadvantage. It is well known that automotive brands overcome their lack of information by demanding access to suppliers' ERP systems.

Artificial intelligence eats transaction costs for breakfast

Company size is usually measured by the number of employees. If economies of scale were the only factor, why are there only a handful of companies worldwide with more than one million employees? At least in China, there should be companies with more than 5 million employees.

With fully automated companies such as Booking.com, Uber or Google Search, there are no longer any employees involved in business transactions. With full automation, transaction costs are close to zero. These are transaction giants and employee dwarfs.

Hypothesis 1: AI leads to the overcoming of industrial work

Automation has been causing the relative number of employees in the manufacturing industry in relation to value added to continue to fall for years. This is not surprising, as the use of capital promotes automation. This reduces the number of employees per transaction.

The fact that unemployment is falling despite automation is due to the rapidly increasing number of relatively small companies and self-employed people. Attentive observers have not failed to notice that the number of mostly small companies in the personal sphere is increasing. This effect has only started to unfold since the introduction of ERP systems and other forms of digitalization.

Our great-grandparents had to do inhumane work in the factories of the industrial revolution because there were no robots, digitalization or AI back then. Today we know: In the long term, this work was not intended for humans at all.

On the one hand, more and more AI machines and robots are being used in factories. In addition to this bottom-up transformation, more and more processes are being integrated top-down into smarter ERP systems until everything is automated in the end. When all the inhuman drudgery has been digitized, what remains is meaningful human work. And we will never run out of this work.

Hypothesis 2: Diseconomies of scale make SMEs the winners of the "AIization of the economy"

The highly efficient coordination of multitudes of small businesses through networked ERP systems has the potential to compete with large companies in the market. We are already familiar with such structures from the clothing industry, where companies such as Nike essentially coordinate marketing and design and outsource everything else. In chip production, too, Intel, Nvidia and ARM only design their products and outsource most of the production to TSMC in Taiwan. Measured against its position as the undisputed market leader in global high-end chip production, the Taiwanese company is relatively small with only 60,000 employees - and the trend is downwards. In production itself, the employees' tasks are now mostly limited to troubleshooting and repairs.

The larger a company, the more human energy is lost due to company politics and other diseconomies of scale. The smaller a company, the higher the proportion of employees working with customers. And in an artificial world, nothing will be more important than human relationships.

Hypothesis 3: ERP becomes the "on-board computer" of the self-driving factory

Language models such as GPT can not only convert text into images, but also vice versa. This means that Google's harmless-looking robot RT-2, for example, is able to perform tasks for which it has not been trained or programmed. This is revolutionary.

The increased use of such smart robots in collaboration with IoT sensors and AI is giving ERP systems an ever better understanding of their own environment. The potential for autonomous production is growing from this increasing global knowledge.

What plant manager wouldn't be pleased to have a friendly on-board computer like the one Captain Picard will have on his starship Enterprise in the distant future?

‍