An ERP was facing a huge problem…

The growing economy causes exponentially more paperwork for companies, as a result, the unimaginable quantity of invoices needs to be integrated into digital databases. It requires a lot of stressful work from people to make this protracted, expansive process without mistakes.

Tried to automate it…

ERP has a bad experience with automation, because the last solution’s properties were 10 minutes/invoice and 2-week training time for a new format. Every automation which inner logic was built by people (loops and conditions) aren’t satisfying in terms of accuracy and processing speed metrics. There are such problems which seem to be easy for us, for humans, but the modern science doesn’t have clue about how it works in our brain yet (optical character recognition, natural language processing, etc…

Finally, they found a proper solution…

However, nobody knows its inner logic, but was born after BigData, and was named as artificial intelligence. ERP trained the AI with 1500 training examples, which has learned to solve the registration tasks: reads, interprets and provides informations. It is solving never seen invoices within ~7 seconds, in a completely format independent way, with 98% accuracy, every day of the week, in every hour. Its operation doesn’t require particularly expensive hardware, so the innovative Hungarian company acquired 300 times better solution than the previous ones.


Time is the real problem of companies…

Our case study partner is an office, with more than 10 years of economic experience. The amount of data coming to their customers is huge, and manage this process requires too many human resources. As the numbers grew, inaccuracy and chaos are increasing. It requires solution as soon as possible.

Conventional solution…

Initially, experts and a systematic program were used to ensure a higher degree of order and supervision. But this solution, unfortunately, meant work time wasting of high-quality professionals.

Now they have just become unstoppable, because…

The perfect solution is to create a completely new assistant. This “colleague” was embodied in the form of artificial intelligence. The structural and semantic based interpretation (with AI) enabled to forward the documents to the competent person, automatically and independently of their extension (be it .DOC, .DOCX, .LOG, .MSG, .ODT, .PAGES, .RTF, .TEX, .TXT, .WPD, .WPS, .CSV, .DAT, .GED, .KEY, .PPS, .PPT, .PPTX, .SDF, .TAR, .VCF, .XML). The system’s performance is ~ 5-23 seconds per incoming data, with 86% efficiency. It runs 24/7, and its operation doesn’t require particularly expensive hardware. ALBA has come a long way, but it’s worth it because their customers are able to enjoy their state-of-the-art services quality experience.



Human factors such as subjective opinion, inaccuracy, inattention, and fatigue…

The connection between the product and its primary material is a series of production processes. Controlling the quality of the raw material is necessary for a good quality product. However, in these quality control processes, the human factors such as subjective opinion, inaccuracy, inattention, and fatigue play a big role. One of these can cause customer complaint and high, sometimes more than 60%, waste of production.

More eyes are able to see more… conventional machine vision

Initially, our partner doubled the human resources working on the raw material which just resulted in slight improvement. The growth didn’t meet with the expectations: the loss continued to be around 50%, and the rate of decline in customer complaints wasn’t convincing.

They tried to rely on conventional machine vision solutions, but the less regular, “intuitive” solutions required more than that. (Identification of contamination, damage, scratches, cracking, with unpredictable size and appearance, etc.). Therefore, any automated solution that has been developed by man from a series of conditions and cycles still leaves a lot of room for improvement speed and accuracy.

“To convince them, you have to offer at least 10 times better …” Elon Musk

The innovative spark came from a young engineer’s mind, who saw the opportunity and the weaknesses of the outdated solutions. The solution was the electricity of the 21st century, the artificial intelligence. Our factory-installed system scans and evaluates the raw material within 3 seconds, with 25-class classification and segmentation at the same time. Recognizes the type of injuries found and determines its shape, size, and outline. Compared to its human predecessors, it reduced the loss to 37%, resulting an impressive ROI for its users.


Failure of critical equipment…

More than 500 machines assist throughout our client’s production process. The working pace of the employees as a unified organization is awesome, machines (winders, fiber machines, industrial scanners, steam presses, etc.) and people together, in sync. The deadline is close: a project followed by the another one. However, the story didn’t always go smoothly, because of some unexpected hidden errors. Sometimes such incident held back the entire production for a long time, which caused a serious loss for our partner.

Routine checks…

Routine inspections with high frequency have been introduced to minimize the number of unexpected failures. It required highly professional, expensive labor from a hard-to-find market. Obviously, the lack of experts and the need for resources from the other side were too high to put the planned routine tests into practice to reach the expectations.

A factory produces data in every moment, but who uses them…?

Our predictive maintenance solution keeps the data (generated by the machines) monitored 24/7. The special advantage of our system is that it also receives inputs like the sound and vibration frequency of the machines, with the usage of installed IOT devices. Thus, if a device stops 2 times between 3 and 4 o’clock in the morning, our system will be aware of it and interprets the upcoming failure events and creates a report. So, it prevents failure before it could happen. Of course, this was an oversimplified example, because through BigData our system is able to detect signs and patterns which are hidden for professionals and thus making warnings.

A predictive maintenance program offers a wide range of benefits to manufacturers. Its first advantage is that it prevents unexpected downtime caused by unexpected errors, which saves economic costs. Secondary quality benefits:

              Reduced outage by 50% due to equipment failure;

              30% increase in average machine failure time (MeanTimeBetweenFailures);

              Brought a 30% reduction in the spare parts kit;

              Reduced maintenance costs by 10-40%;

              Product quality improved and customer satisfaction increased;

              10-20% reduction in waste.

As a result, our partner spent less time with being passive, spending less money on physical maintenance, and requiring less human effort to handle all of the supervision.

Can we see in the future? Demand forecast? Yes.

Warehousing, receipt, supply and demand, cost of overstock and inventory shortages. The concepts listed above create and rule the existence or failure of businesses. Simple, because the loss isn’t an elusive cloud, but it is perfectly evident, how much more you threw out of the window. How many product would you keep in stock if you know the exact demand for the future?

They shouted: The key is statistics (or not)!

We may think that statistics can be a strong point of reference for our revenue strategy. But it doesn’t see in the future. It surely prevents you to being lost and keeps you within a wide interval. Even with excellent statistical models, such as autoregressive and/or moving average models, or conditional distribution-based models, near-perfect intake can only be achieved if your labor have superior data processing capabilities in the detection of trends and patterns.

Recognition of patterns and hidden indicators in demand data series?

It’s possible to forecast the demand product by product. According to our partner, they receive such forecasts from our software, which is unimaginable. Awesome and at the same time brilliant. There are more than 15,000 different products in their dealership, for some they have already an automatic demand forecast, which makes their inventories and receipts to be more optimal than ever, which fact was proven with the past data of our partner.


Agile scrum development methodology analyzer scrum master advisor based on artificial intelligence.

Typically, codebeamer clients are experiencing development projects that, due to the nature of the development, include unexpected events that require the use of agile methodology. The incorrect application of the procedure leads to chaos in a crowded environment, which has lack of labor. Scrum masters who lead the projects must have high-quality skills to closing their projects in an organized way, without any misunderstanding. Outstandingly difficult tasks, misunderstandings and lack of transparency often lead to overdue and distrust of customers.

Labor shortages …

The plan is to introducing more experts, but there are only a few highly trained freelance engineers on the market who could save the ongoing several 10,000 companies in a few months from deep water.


The circle is continuously narrowed down, technology left as a lifeguard of humanity, typically in such situations. Which seems to be chaotic for people, in the eyes of the electricity of the 21st century (AI), a large amount of data, is a playground. Our development, an artificial intelligence-based scrum master advisor. Observes the scrum master’s projects and teams, and analyzes each team. In its analysis, it isn’t only compares a few metrics and curves but also processes hundreds of data series in seconds, filtering out clear relationships and indicators of failure. As a result of the analysis, it will decide how much the project is at risk and immediately notify the scrum master of what has happened. Obviously, the system is running automatically by itself on-the-fly.

Copyright © 2019 AiCAN All rights reserved.

Close Menu