Revolution in quality management or just evolutionary development?
Artificial intelligence is now on everyone’s lips. The widespread publication of trend-setting systems such as ChatGPT and DALL-E has not only made this technology of the future accessible to a broad mass, but also lifted it out of the hidden depths of modern software engineering. Artificial intelligence (AI) is suddenly no longer something abstract hidden behind spectacular filters in photo and video software or chat bots that appear human, but a system with which one can communicate directly and freely. Although the technology has not really been tested for long, it is already astonishing to many how difficult it is to distinguish today’s AI systems from a “real” human being in direct dialogue – indirectly, one sometimes feels reminded of the classic film “Bladerunner”, in which only the evaluation of a huge amount of questions can identify artificial persons at all. But we haven’t quite got that far yet.
What actually constitutes artificial intelligence?
The transition to systems based on artificial intelligence marks a milestone for the technical development of humanity in many ways. In purely technical terms, it marks the transition from an era of rule-based software development to one based on learning. Systems are no longer pure, deterministic flow charts, but their actions and responses are based on learned and combined information. They are thus similar to human information processing, which can come to different results in “identical” situations.
From an economic point of view, the great hopes for artificial intelligence lie, of course, in the area of further increases in efficiency. The often expensive human factor, according to the goal, should be replaced by intelligent systems wherever possible and justifiable, and should also decide and act autonomously as far as possible during operation. Artificial intelligence should also bring improvements where humans can cause problems through errors and inadequacies. The hoped-for goals are thus very similar to the expectations already placed in general digitalisation, which, as we know, have not always come about on their own.
From a social point of view, AI marks a change in communication. For the first time, it is possible to communicate with the machine in a truly general and literally “colloquial” way. Control with gestures will also increasingly find its way in here. User interfaces will thus become dramatically less important and the circle of people who can participate in communication at all will expand drastically. Incorrect or misleading inputs to systems can be corrected or completed automatically.
Application areas for AI in quality management
AI-based approaches have great potential to make quality management more efficient and precise.
- Quality control: AI-supported image processing systems can analyse products in real time and detect quality deviations. This reduces human error and ensures consistently high product quality. Feedback to production can thus also reduce waste and optimise the need for deployed work equipment. Automated checks or virtual assistants can optimally support the human user.
- Process optimisation: AI can be used to monitor and analyse complex production and business processes. By identifying bottlenecks and inefficient workflows, companies can optimise their processes and increase overall quality. This increases not only quality but also throughput, if necessary, and offers good approaches for a more demand-oriented use of production resources.
- Predictive maintenance: By analysing sensor data, AI algorithms can detect patterns, irregularities and anomalies that indicate an upcoming need for maintenance or updates. This enables companies to perform machine and system maintenance in a timely manner, thereby minimising unplanned downtime.
- Supplier evaluation: AI can help evaluate supplier performance by analysing data from various sources and providing indications of supplier risks or quality problems. It can independently order or reduce goods and services according to demand.
- Customer feedback analysis: By analysing customer ratings, comments and social media, companies can gain deeper insights into customer satisfaction and take targeted action to improve quality.
- Creation of documentation and training materials: Digitisation in general and AI in particular are excellent for automating and, above all, updating time-consuming documentation tasks.
Advantages of the use of AI in quality management
Integrating AI into existing quality management offers numerous, obvious benefits:
- Increased efficiency: Artificial intelligence can automate especially those tasks that are normally time or resource consuming, resulting in increased efficiency and resource savings.
- Real-time analytics: AI enables real-time analytics that allow organisations to respond quickly to quality issues and take preventative action. The risk of dragging undetected problems into the production process for a longer period of time is drastically reduced. Automated risk assessments can also detect slowly developing problems at an early stage and take appropriate countermeasures.
- Precision: AI-based systems are able to recognise patterns and process complex data, which leads to more precise results – provided the learning tools used are correspondingly good (see below: “Bias and fairness”). The ability to learn independently ensures that processes can automatically improve themselves over time. Particularly interesting approaches to this are already frequently found today in the automated selection and determination of samples, for example.
- Cost reduction: By minimising waste, errors and downtime, companies can reduce costs and increase profitability. Efficient structures also allow lower end prices, which can be the decisive factor for business success in today’s global competition.
Risks and "side effects" using AI
Besides the many and undeniable benefits, there are also risks and challenges in the context of using AI – not only in quality management:
- Data protection and security: Processing sensitive company data requires robust security measures to prevent unauthorised access and data leaks. End-to-end automation may also create new single-point-of-failure structures that require appropriate safeguards. The need for highly specialised – and, of course, expensive – skilled staff for the entire system and IT area inevitably increases, unless one wants to rely on external forces.
- Human supervision: It remains important that human experts monitor and interpret the results generated by AI systems in order to avoid “automated wrong decisions”. The responsibility of these experts is correspondingly high, because they must also be able to take tough measures in case of doubt. Moreover, the employees need extended training that also goes beyond their actual core competence.
- Bias, fairness and ethics: AI algorithms could develop unintended “biases” or “preferences” based on training data, which could lead to biased, unfair or even discriminatory results. Target groups or target markets could gain too much influence on certain factors under certain circumstances, leading to undesirable polarisation. The “learning tools” of AI must thus not only be considered from a purely technical and economic level, but may also require ethical, moral and social evaluation. Ultimately, even a possible “question of liability and guilt” must be clearly answerable. Who is actually responsible when AI makes “wrong” decisions? The programmer, the user, the operator, …or the AI itself?
- Complexity and implementation: The introduction of AI requires very specialised knowledge and resources. Smaller companies in particular may find it difficult to meet these challenges and fall behind as a result, despite having good products. In particular, the ever-increasing complexity – and thus issues such as maintenance and updating – pose a risk that should not be underestimated.
“Bias, Fairness and Ethics” – The BIG Problem of Artificial Intelligence
The topic of bias requires special attention when considering the disadvantages and problems of AI. It is not yet clear whether it will finally be possible to make systems REALLY similar to human intelligence. This is particularly noticeable in areas where terms have multiple meanings, possibly in foreign languages or even with different spellings. In other words, where a human brain can specifically differentiate on the basis of its experience and expertise.
A prominent example is the case of the social network Facebook. Several service providers wanted to post training courses in Facebook’s own advertising – specifically on the topic of “Python” (programming language) or a Python library for data evaluation and visualisation called “Pandas”. The quite obviously AI-controlled ad review evaluated the offers as prohibited animal trade and reacted rigorously – with lifelong blocking of the corresponding users. What is particularly piquant about this is that even the built-in complaint function of the Facebook ad management probably led to an automated and AI-controlled response to the inquirers in many cases. A very good, but also frightening example of what is possible if
- an AI is given sole freedom of choice
- learning material is obviously used unilaterally to train an AI (here: animal trade) without taking into account, clearly delineating or perhaps even wrongly prioritising double concepts
- the human factor has (for unknown reasons) even been excluded from revision processes.
Now, a rejected advertisement or even a lifelong Facebook block is ultimately only annoying for the individual concerned and perhaps also partly economically damaging, but similar mistakes can – and of course will – also happen in more sensitive areas. At the latest when AI enters areas that involve physical integrity (care, medicine, traffic,…), there will be no way around putting all the successes achieved so far to the test again.
AI in quality management - an assessment
Artificial intelligence has the potential to revolutionise quality management (a Game Changer) by increasing efficiency, precision and responsiveness. Companies that skilfully use AI in quality management can benefit from improved product and service quality. However, the challenges and concerns around privacy, human monitoring, bias and implementation should be carefully addressed.
Similar to digitalisation in general, AI should not be misunderstood as a universal problem solver or saviour. It is also a tool that can bring many competitive advantages when used in a targeted manner, as long as the risks are taken into account accordingly. The future of quality management will undoubtedly be shaped by the intelligent integration of AI.
By loading this video you agree to the privacy policy of YouTube.
Read more
WhitePaper: “Remote Audit – Virtuelle Fern-Audits” von Dr. Roland Scherb (PDF, ca. 1.15MB)
AI and Quality Management.
We can talk about it!
Contact us now.