What is “artificial intelligence”?
“Artificial intelligence” refers to the ability of machines to mimic human behavior as deceptively as possible. Already, AI applications are extremely powerful, and further development is exponential. Applications have been in the works for several decades. However, the public only became aware of AI capabilities in late 2022 with the first publicly available version of ChatGPT.
Since early 2023, applications like ChatGPT have been the talk of the town. But what does “GPT” actually mean? GPT stands for “Generative Pre-trained Transformer”. And that already tells us what we can expect from an AI of this generation: It must be trained to deliver results.
Various methods are suitable for this training. Data-based learning methods are particularly suitable for logical connections; expert knowledge can be captured well with rule-based methods. However, these methods can only reproduce known relationships. If essential framework conditions change, these experience-based systems would have to enter new territory themselves. Therefore, they will not be able to provide reliable answers in such situations.
ChatGPT has been trained to run dialogs conducted in natural language. It is based on a large amount of data with already conducted arguments about certain topics, taken from many sources. Human feedback on the initial output of AI applications like ChatGPT has further trained and shaped the application and will continue to make it better and better in the future. This reinforcing learning through use is leading to generative language models that can now engage in human-like conversation. In many cases, the machines that identify themselves as conversational partners can no longer be distinguished from humans in their speech. What’s more, they can even visually present themselves as naturally rendered moving avatars that are virtually indistinguishable from humans.
How is artificial intelligence impacting businesses?
The availability of AI applications will have a strong impact on the vast majority of industries and businesses.
Generative AI applications can be used in all areas of activity to increase efficiency. These mainly include repetitive administrative, testing and interaction tasks. However, it has also been shown that AI applications are even capable of taking over intellectually demanding tasks that were previously performed by highly skilled professionals. For example, planning, programming, and even creative tasks are already being performed amazingly well by AI applications.
In times when many jobs cannot be filled by qualified professionals, AI applications are not inconvenient for companies. Companies should therefore take a close look at the opportunities AI applications offer.
Decision support through AI – Which decisions can be supported by AI or even transferred to AI applications?
AI applications can be trained to be able to filter which complaints are likely to be justified and which are likely to be unjustified. Employees can then focus on the more interesting, uncertain cases, better target their skills, and enrich the value of their jobs.
In bakery factories with connected sales outlets, AI applications can generate recommendations on the production quantity of baked goods that makes sense in each case and thus achieve an optimum of delivery capability and the quantity of unsold goods. In their analysis, AI applications incorporate historical sales data as well as weather data and data on special local events.
Farmers using AI for irrigation control have been shown to benefit from yield increases and water cost savings simultaneously.
Companies using AI applications for decision-making can reduce their operational business risks and better manage high staff turnover. It’s understandable, of course, that experienced experts are reluctant to cede decision-making authority to a machine. Companies should focus on a fruitful interaction between machine-generated suggestions received and expert know-how. This allows AI applications to be better trained and experts to gain confidence in AI recommendations.
Integrating AI into customer interactions: What standard interactions can AI applications perform?
Chatbots can already perform standard interactions with customers well and safely. The possibilities already go far beyond computer telephony integration (CTI) applications. Customer surveys and customer communication for inquiry and order processing can be supported by AI applications or even executed independently, not only in writing but also via natural language.
More complicated cases can be handled by (human) employees. Attention should focus on the point of handoff between machine and human. In well-designed systems, customers do not even notice the transition.
There is also a dynamic aspect: as the capabilities of AI applications increase, the boundaries are in fact shifting further and further in the direction of humans. To keep efficiency at the highest reasonable level in each case, it is advisable to continuously review these boundaries of interaction systems and adjust them if necessary.
Processing through AI
What kind of factual processing can be performed by AI applications?
Programmers already have runnable and executable code created for delimited routines by AI applications. The time required to brief an AI application with a prompt is significantly shorter than the time required for human programming.
AI applications can also perform design tasks in mechanical and structural engineering. Briefed in a targeted manner, AI applications can design out modules and save designers time. The work of designers will shift in part to writing instructions to AI applications and evaluating and reworking their design proposals.
Following on from the example of complaints, AI applications can write response letters in which they refer to the customer’s letters and either agree to warranty services or reject them with justification.
Companies that use AI applications in processing become more efficient and improve the quality of their employees’ work. The prerequisite is appropriate training for employees to be able to use AI applications in a meaningful way.
Creativity with AI
Which creative processes can be supported by AI applications?
Creativity is required in writing texts as well as in composing or in art. Unfortunately, creativity is not always retrievable. The suggestions that AI applications generate can provide impulses that can then be reworked by employees and further executed with their own touch. This shortens creative processes and makes them more efficient. Employees work in dialog with an AI application instead of being left to their own devices.
This will change the way journalists, copywriters, graphic designers, musicians and other professionals work. Those who do not use AI applications in the future will likely lose relative efficiency and suffer competitive disadvantages.
How should professionals, managers and companies adapt to AI?
Will AI replace us?
The big fear is that AI applications will substitute jobs. These fears are understandable given the increasing capabilities of AI applications. However, if we look at the introduction of computers into workplaces and the introduction of robotics into manufacturing, we see that these revolutionary technologies have not led to job losses. They have, however, brought with them a reorganization of skills. Personal computing has brought with it new requirements needed to set up and maintain computers and computer systems. The introduction of industrial robots has raised requirements related to programming (teaching) and maintenance of these robotic systems. It will be similar with the introduction of artificial intelligence in companies. The handling of these new applications must be learned. Accordingly, companies should offer their employees suitable training programs. Specialists and managers should not suppress the new technical possibilities, but should get to grips with them in order to safeguard their employability.
What can AI do for us?
Instead of surrendering to AI, consider what AI can do for your business. In particular, AI applications enable people to relieve themselves of routine work and to increasingly assume their role as thinking, creative beings. When AI is used to free up human capacity for challenging tasks, employees and companies can benefit. The fields in which AI can provide concrete benefits in the company in the future must be identified as part of an AI strategy. Pilot projects can then be used to deploy selected AI applications in a targeted manner.
The introduction of AI has an impact on the way we work. That is precisely the intention. However, the interaction between humans and machines must be redefined and practiced. Cooperation must first become established. During this transitional period, managers need to be tolerant and supportive.
AI can make people’s work easier and relieve them of standard tasks. The decision-making competence of AI applications described above, combined with execution competence, allows triage to be implemented. AI takes care of clear cases, while doubtful, more challenging cases are solved by employees.
Given the large generational shift, AI can also help companies retain experiential knowledge that resides in lived processes. For this purpose, knowledge should be absorbed by AI as long as it can be retrieved by employees. If this transfer of factual and process knowledge is transferred well to intelligent systems, it doesn’t matter if hints and recommendations come from machines instead of people in the future – the main thing is that the knowledge is available.
The experience that companies are in danger of losing due to increasing turnover in simple functions can also be absorbed by AI. AI applications can provide valuable assistance to new and inexperienced employees in their jobs and help ensure that processes run effectively and efficiently regardless of individuals.
Step by step, companies can also incorporate AI applications into creative tasks. AI-generated impulses and proposed solutions can be taken up by employees, meaningfully adapted, supplemented and rounded off.
We are not at the mercy of artificial intelligence, but can shape ourselves what we want to use AI for and how. This is where the real challenge lies: recognizing the opportunities to deal with AI constructively and critically reflexively at the same time, and to promote our own ability to think and solve problems.
What should companies pay particular attention to when using AI applications?
Strategic and operational aspects when using AI applications
The technology of self-learning AI can create problems if large portions of the available information are generated with AI applications. This is because the applications then increasingly draw on their own previous output. This can lead to self-reinforcement effects. The generated results then threaten to move further and further away from reality. Therefore, a high degree of judgment is required for AI-generated results. In particular, if AI applications do not provide the sources of information, users should carefully check the information offered for its truth content.
Another aspect is significant: the development of AI technology and the majority of the data used by AI is largely in the hands of a few companies, all of which are US-based: Microsoft with OpenAI, Amazon, Alphabet with Google, as well as Facebook. The training of AI applications is also largely under the direction of US companies. Thus, it can be assumed that the results are shaped accordingly. Users should know this when they use AI systems.
Because the quality of the results that generative AI applications can produce depends centrally on the quality of data available to them, digitization is a necessary first step on the road to artificial intelligence. In particular, it depends on the availability of comprehensive, good data in the specific application domain of enterprises.
AI applications can apply learned patterns to this database to produce results. If the data is publicly available, virtually anyone could generate similar results. Competitive advantages in the future will therefore come from having exclusive large and relevant data sets. Companies should therefore definitely make their employees aware of the value of data.
Further competitive advantages will be tapped by the special quality of the query of the performance. The formulation of so-called prompts will influence the quality of the results that an AI application produces. In-house data resources and in-house prompting expertise will therefore have a decisive impact on the competitiveness of companies.
In any case, companies should develop and pursue an integrated digitization strategy that includes the requirements for using artificial intelligence. The digitization strategy is a prerequisite for an AI strategy based on it.
Legal aspects of using AI applications
Legal issues arise in the development, parameterization and use of AI applications.
If you create AI applications for third parties or have AI applications created by third parties, this development service is performed on the basis of a contract for work and services, in which the services and liability for defects must be regulated. To avoid problems, both the performance and the intended use should be specified in concrete terms. In particular, it should be agreed whether the developer or the customer is to train the AI application. To prevent the phenomenon of machine bias, the requirements for the quality of the training data should also be agreed as precisely as possible. It must be excluded that the AI application with its results violates laws and guidelines, e.g., discrimination prohibitions. With regard to the training data, the protection of personal data must also be ensured. Clients who obtain training data from third parties should obtain confirmation of compliance with these requirements. It is also frequently recommended that clients should have the AI developers certify traceability. However, this will hardly be possible in practice due to the complex decision-making processes and the learning ability of AI applications. Users of AI applications must comply with sector-specific regulations that apply, for example, to applications in healthcare, road transport or financial markets. Clients should obtain confirmation from their developers that the applicable regulations are being complied with. For example, affected parties must be informed when decisions are made by an AI application; in some cases, it is even prohibited for decisions to be made solely by an AI application (lending).
The quality of training and training data significantly shapes the quality of AI applications. Proven training data is correspondingly valuable. Therefore, it is essential to agree who will be the owner of the valuable training data after successful parameterization.
While software can be copyrighted, algorithms and parameterizations cannot. It is important to know that AI application can hardly be protected by copyright or patent because they are based on algorithms and their parameterization. Only the factual lead in the market is eligible for protection.
The still unresolved legal question of who owns the results creatively produced by AI applications (images, texts, films, constructions, etc.) is also interesting. Legal certainty can only be achieved through a binding and enforceable agreement between the developer of the AI application, the supplier of the training data, the trainer and the company using the AI application.
There is another aspect: if you want to integrate AI applications into smart products that are covered by product liability, that product liability also extends to the AI applications you incorporate. Companies that use AI applications are fully responsible for the consequences of that use. AI applications, even if they should make autonomous decisions, are not legally independent entities. Responsibility and liability always lie with the company using the AI application. Traceability of the decisions provided by AI applications is usually not possible due to the complexity of the decision-making processes. Companies can only escape liability if they can prove that errors that led to harm to third parties can be traced back to errors in the neural network or to training data or the training itself that service providers provided or performed, or that the AI application was used for an unintended purpose.
Until legislation has conclusively addressed the use of AI applications, companies should take their own organizational du contractual measures to deal with the legal risks.