Why companies find it so difficult to use AI

Konrad Wolfenstein

1 year ago

Why companies find it so difficult to use AI – Image: Xpert.Digital

Leveraging AI potential: Strategies for the companies of tomorrow

AI in business: Challenges, solutions and future perspectives

The rapid development of artificial intelligence (AI) has created a multitude of opportunities for companies in recent years. AI can, among other things, automate processes, analyze data, generate forecasts, support employees, and open up entirely new business models. Despite these promising prospects, many companies still struggle to profitably integrate AI applications into their operations. Often, they lack the technological foundations, the necessary expertise, and a corporate culture open enough to the associated changes. Added to this are legal and ethical concerns, as well as uncertainty about how AI will affect jobs and organizational structures in the long term. This article highlights the key challenges, identifies success factors to help companies overcome these hurdles, and provides an outlook on the future of AI in business.

1. The main obstacles to the introduction of AI

Technological complexity and integration

AI systems are often based on complex machine learning algorithms that require a robust IT infrastructure and highly specific knowledge in areas such as data science, software development, and statistics. A major hurdle is usually adapting and, if necessary, restructuring existing databases, ERP systems, or other software solutions. In many cases, companies even have to implement entirely new platforms or interfaces so that the AI models can access the necessary information.

Another challenge is the shortage of qualified specialists. While interest in data science, machine learning, and AI is growing, the demand within companies often outpaces the training and development opportunities for experts in this field. Even when companies actively seek out talented AI specialists, finding them and successfully integrating them into the organization is not always easy. One approach is to offer in-house training programs, provide further training for existing employees, or utilize external consulting services. Some companies are exploring practical, innovative approaches to fill knowledge gaps through collaborations with universities or startups.

Data security and data protection

AI applications typically require large amounts of data, which, depending on the use case, may contain sensitive or personal information. This places high demands on data security and privacy. Companies must implement technical, organizational, and legal measures to ensure that personal data is not misused and that all relevant data protection regulations are complied with. For example, when AI systems are used for forecasting, recommendations, or automated decision-making, the likelihood of sensitive data being aggregated and processed on a significant scale increases.

Compliance with legal requirements and international standards is only one side of the coin. Equally important is strengthening the trust of customers, partners, and employees in AI solutions. A professional approach to data quality and data integrity is crucial in this regard. AI models trained with faulty or manipulated data deliver unreliable, and sometimes even harmful, results. Therefore, it is essential to establish appropriate security protocols that, for example, protect against unauthorized access and data manipulation. Even a single data leak can permanently damage a company's reputation and seriously jeopardize an AI project.

Liability for damages

A particularly important issue to consider in AI applications is liability. What happens, for example, if an AI-controlled device or system causes damage? Take the self-driving car: If it injures pedestrians or causes an accident with other road users, companies or courts must determine whether the vehicle owner, the software developer, or the manufacturer is responsible. The legal situation in this area is still evolving worldwide, as it is a relatively new field in which laws, norms, and standards are only gradually being developed and defined.

Furthermore, additional questions arise: If their AI systems malfunction, are development teams or companies required to demonstrate precisely how a decision was reached? Is there an obligation to disclose the AI algorithm to clearly identify which part of the process led to the error? Such aspects demonstrate that the AI industry is characterized not only by technical complexity but also by legal uncertainties. Companies should therefore address potential liability risks early on and stay informed about legal developments in the field of AI.

Change management and cultural acceptance

The introduction of AI technologies often means a fundamental change in a company's workflows and processes. Employees have to adapt to new tools, software solutions, and ways of working. It's not uncommon for fears to circulate that AI systems will completely replace human tasks or that work will be more closely monitored. This leads to resistance to change, especially when employees cannot understand the purpose and benefits of the new technology for the company and for themselves.

The willingness to admit mistakes and learn from them is a key element in dealing with AI. Algorithms don't function flawlessly from the outset. They often need to be iteratively trained and optimized until they deliver reliable results. An open culture of learning from mistakes, where new ideas and experiments are encouraged, fosters acceptance. Furthermore, leadership plays a crucial role. If the executive team or management initially supports an AI project enthusiastically but then loses interest, it can unsettle employees. Continuous engagement and regular performance reviews by top management help increase the acceptance of AI throughout the entire company.

Cost and resource management

AI projects can be very costly. Not only does the acquisition of the technology incur high expenses; companies also need suitable hardware infrastructure (e.g., high-performance servers), must license software solutions, and build data platforms. A significant portion of the budget can also be allocated to employee training or collaboration with external AI specialists.

At the same time, successfully implemented AI solutions often offer considerable added value. They increase productivity, accelerate workflows, and reduce operating costs in the long term. Therefore, defining measurable goals and key performance indicators (KPIs) is essential when considering the cost-benefit ratio. Companies should not only ask what specific added value AI creates, but also how quickly the investment will pay for itself. In some cases, it may be economically advantageous to initially rely on standardized AI solutions or cloud-based services instead of commissioning expensive, custom-developed solutions. In other situations, however, a custom-programmed AI – for example, for highly specialized industrial applications – may be the best solution.

Ethical and legal challenges

AI systems can make decisions automatically or at least strongly influence them. This creates a responsibility to examine these systems for fairness, transparency, and non-discrimination. If AI models are trained with biased datasets, they could systematically disadvantage people or draw incorrect conclusions. Ethical questions surrounding surveillance, facial recognition, emotion recognition, and the intrusion into privacy are also becoming increasingly prominent in this context.

In many countries, governments, associations, and expert panels are discussing regulations to ensure that AI remains trustworthy and serves humanity. A growing number of companies are developing their own AI ethics guidelines to be perceived as responsible and to avoid potential scandals arising from discriminatory or opaque AI practices. This ongoing debate demonstrates that the issue is not only technically relevant, but also socially and politically.

2. Success factors for a successful AI implementation

Despite the aforementioned obstacles, numerous companies are already successfully using AI in their processes and products. Their experiences offer valuable insights that can serve as a guide for other organizations.

Clear objectives and strategy

A precise definition of goals is the starting point for any successful AI project. Companies should ask themselves in advance which specific problems or challenges they want to solve with the help of AI. An AI project that is not focused on clear use cases risks having unclear benefits or making them difficult to measure.

The AI strategy should also be integrated into the overall corporate strategy. This requires a shared understanding of how AI enhances innovation, enables new products, or makes business processes more efficient. Such integration ensures that the relevant business units and departments are involved in the planning and that the necessary resources are available in the long term.

Data management and quality

Data quality is a crucial factor for the performance of AI. For machine learning to be used effectively, extensive and, above all, clean datasets are required. Even collecting relevant data can be complex, especially when different departments or subsidiaries store their information in isolated systems.

Professional data management includes data preparation and cleansing. Poor data quality can lead to inaccurate forecasts, misleading insights, and financial losses. Many companies therefore invest in data infrastructure, data integration, and data governance. A central data platform used by all departments also improves collaboration and enables a consistent understanding of the data across the entire organization.

Interdisciplinary teams and agile methods

An AI project is rarely just the IT department's responsibility. Success requires collaboration among professionals from diverse disciplines: data scientists, software developers, subject matter experts from the affected business unit, UX designers, project managers, and often also lawyers or ethics experts. Connecting these different roles leads to a more comprehensive view of the problem and enables creative approaches to finding solutions.

Agile working methods like Scrum or Kanban are particularly suitable because AI projects are typically carried out iteratively. A model is trained, tested, adapted, and retrained – this cycle repeats frequently. Rigid project planning, where every step is defined down to the smallest detail in advance, is less appropriate. Iterative phases and regular feedback ensure that errors can be identified and corrected early on. Furthermore, new insights can be continuously incorporated into the project.

Continuous monitoring and adaptation

AI models do not automatically remain accurate and efficient indefinitely. If the environment changes, for example due to new data sources, differing customer needs, or altered market conditions, it may become necessary to adapt or retrain the model. Therefore, it is advisable to establish processes within the company that enable continuous monitoring of AI systems and their performance.

Such processes can include meaningful key performance indicators (KPIs) for measuring the success of AI implementation. If deviations are detected, the team must react promptly. This ensures the AI solution remains up-to-date and retains its practical relevance. Furthermore, monitoring is a fundamental aspect of quality assurance, preventing incorrect decisions or systematic biases that might only become apparent after some time.

Training and further education

A new technology will only successfully take root in an organization if employees are empowered to use it. This applies to managers, who need to understand the strategic importance of AI, as well as to specialists in the affected departments. Depending on the use case, some employees only need an introduction to the basic principles of AI, while others require intensive training in specific algorithms, programming languages, or machine learning methods.

Suitable training and development programs not only increase efficiency in the application of new tools and processes, but also strengthen acceptance. Those who are given the opportunity to develop their skills and learn new things are more likely to perceive technology as an opportunity than a threat. From a company perspective, investing in such programs is worthwhile because it builds internal expertise that is essential for future innovation projects or complex AI initiatives.

This goes well with:

Foreign language AI for SMEs: Become a language whiz with artificial intelligence – simple implementation, cost savings and no maintenance required

3. Examples of successful AI implementations

A look at some well-known companies shows how diverse AI can be used:

Amazon: This company makes extensive use of AI, for example for personalized product recommendations or to optimize its supply chain. AI-powered analysis of images and videos also plays a role.
Meta platforms: These platforms use recommendation systems and algorithms to detect unwanted content. The goal is to show users relevant posts while simultaneously curbing the spread of harmful content.
Tesla: In the automotive sector, Tesla uses AI for autonomous driving. The camera and sensor data from its vehicles are constantly analyzed so that the system can learn and, ideally, become increasingly safer.
Upstart: In the financial sector, the company uses AI-powered algorithms to assess the creditworthiness of borrowers. The goal is to make more precise credit decisions and accelerate loan application processes.
Mastercard: Here, AI applications are used, for example, in customer service and fraud prevention. The algorithms help to detect irregular transactions and quickly initiate corrective action.

These examples illustrate that AI is by no means just a topic for tech giants, but is also being successfully used in the financial and insurance sectors, in industry, and in many other sectors. The common denominator lies in a clear definition of goals, excellent data management, and a corporate culture that allows for experimentation with new technologies.

4. Types of AI projects

For a company to successfully implement AI, a fundamental understanding of the different types of AI is helpful. A common distinction is made between weak AI, which specializes in clearly defined tasks, and strong AI, which is intended to one day replicate the full breadth of human intelligence. The latter currently exists only in theory and research, while weak AI is already being used in a great many concrete applications.

Weak AI

Weak AI refers to applications specifically designed to solve particular problems. Examples include chatbots, image recognition software, recommendation algorithms, and voice assistants. These AI systems can achieve impressive results within their assigned tasks—for example, recognizing objects in images or understanding spoken language. However, they are not capable of similar performance outside their narrowly defined area of application. Most solutions currently used in a business context fall into this category.

Powerful AI

Strong AI aims to develop a general, human-like understanding and the ability to learn and solve problems independently. So far, it exists only in the imaginations of researchers and science fiction authors, but the discussion surrounding its potential development is growing. Some experts speculate that one day an artificial intelligence will emerge that improves itself independently and surpasses humans in many cognitive abilities. Whether and when this will happen, however, remains an open question.

Typology according to function

Sometimes AI is also classified according to how it works:

Reactive machines: They only react to direct inputs, without storing memories.
Systems with limited storage capacity: They use past data to derive future decisions. Self-driving cars, for example, can store traffic and sensor data and draw conclusions from it.
Theory of mind: This refers to the ability to understand and respond to human emotions and intentions. Such systems are not yet in practical use, but are the subject of research.
Self-awareness: In this scenario, the AI would develop its own consciousness. This is also still purely theoretical.

5. Employee concerns regarding AI

Skepticism towards new technologies is not a phenomenon limited to AI, but reservations are sometimes particularly pronounced in this area. Some typical concerns include:

Job loss

Many fear that automation could jeopardize their jobs. This concern is particularly prevalent in manufacturing environments or service industries where routine tasks dominate. While AI can indeed take over repetitive activities, it also creates a need for new roles in many cases, such as those involved in the support, maintenance, and further development of AI systems, or in advisory positions.

Changes in working methods

AI can change process flows. Certain steps become obsolete, automated analyses accelerate decision-making, and new tools complement daily work. This often leads to a shift in job profiles, which can cause uncertainty and stress. Many employees initially lack a clear understanding of the specific benefits they themselves will derive from AI and how it can contribute to increased efficiency.

Data protection and surveillance

Also relevant is the potential infringement on privacy. AI tools can collect data on employee behavior, performance, and communication patterns. This raises concerns that management will exert greater control over employees or that sensitive information could fall into the wrong hands. Transparent rules and an open communication culture are particularly important here to avoid misunderstandings.

Dealing with concerns

Companies should take their employees' concerns seriously, listen to them, and work together to find solutions. This can be achieved through regular information sessions, workshops, or training. It's also important to highlight how AI can complement, rather than replace, human work. Those who understand that AI can create new opportunities for creative or more demanding tasks are more likely to support the use of this technology. Clear data protection policies that safeguard personal data also strengthen trust.

6. Ethical Implications of AI

Beyond the technical and economic questions, the use of AI in business and society raises a number of ethical issues.

Distortion and discrimination

AI systems make decisions based on data. If the training data is biased or reflects societal inequalities, the AI system can reproduce these distortions unnoticed. For example, applicants with certain characteristics could be systematically disadvantaged if the AI system considers them less suitable based on historical data. Companies must therefore pay attention to how their algorithms are trained to prevent unconscious discrimination.

Transparency and accountability

Even if an AI model delivers outstanding results, the question remains: how did it achieve them? In complex neural networks, the decision-making processes are often not directly traceable. Companies and authorities are increasingly demanding transparency so that customers, users, or those affected can understand how an AI arrives at its result. Furthermore, it is crucial that, in the event of damage or incorrect decisions, it can be determined who is responsible.

Data protection and privacy

AI systems that analyze personal data exist at the intersection of innovation and privacy. The blending of different data types and increasing computing power make it possible to create detailed profiles of individuals. While this can enable meaningful personalized services, it also carries the risk of surveillance and misuse. Responsible companies therefore define ethical principles that clearly stipulate what may be done with the data and where the boundaries lie.

Social Manipulation

AI can not only process data but also generate content. This creates the risk of disinformation and manipulation. For example, AI can be used to create and disseminate deceptively realistic images, videos, or news stories. Companies' social responsibility increases when their algorithms can contribute to the spread of misinformation. This necessitates thorough review processes, labeling, and internal control mechanisms.

Accuracy and ownership of AI-generated content

The increasing use of AI tools to create texts, images, or other content raises questions about quality and copyright. Who is responsible if AI-generated content contains errors or infringes on the intellectual property of others? Some companies have already experienced having to correct AI-generated articles or reports after the fact. Careful review, a review process, and clear copyright rules can help avoid legal disputes.

Technological singularity

A long-term scenario under discussion is the point at which artificial intelligence surpasses humans in many areas. This so-called moment of "technological singularity" raises fundamental ethical questions: How should we deal with an AI that learns and acts independently? How do we ensure that it respects human values and fundamental rights? While such a powerful AI is not yet a practical issue, the debate surrounding it raises awareness of key principles of control and accountability.

Dealing with ethical challenges

Companies using AI technology can establish their own ethics committees or guidelines. For example, clear protocols for data collection, algorithm development, and testing are necessary. Transparent documentation and regular audits increase trust in the technology. Furthermore, organizations should engage in dialogue with society, such as through discussions with stakeholders or public information events, to identify and address concerns early on.

7. The Future of AI

AI is constantly evolving and will likely become even more deeply embedded in our daily lives and the workplace in the coming years. Some trends are already emerging:

Multimodal AI: Future AI systems will increasingly process data from various sources and in different formats simultaneously, for example, text, image, video, and audio. This will enable more comprehensive analyses and more complex applications.
Democratizing AI: AI tools and platforms are becoming easier to use, giving access to smaller companies and departments without large budgets for development teams. Low-code or no-code solutions are accelerating this trend.
Open and smaller models: While large, proprietary AI models have dominated so far, a trend towards smaller, more efficient, and also open models is emerging in some areas. This allows more organizations to participate in AI developments and build their own solutions.
Automation and robotics: Self-driving vehicles, drones, and robots are becoming increasingly powerful. Once the technological hurdles (e.g., safety, reliability) are overcome, their use in areas such as logistics, production, and service is likely to increase very rapidly.
Regulation: As the importance of AI grows, so does the call for legal frameworks. Future laws and standards will more strongly guide the development and application of AI in order to ensure, for example, security, data protection, and consumer protection.

Impact on the economy

The economic importance of AI is likely to increase further in the coming years. Automation will set new standards in many industries, and companies that successfully adapt to AI early on will gain a clear competitive advantage. At the same time, new business areas are emerging in which startups and established companies can develop innovative applications. Particularly in the areas of data analytics, healthcare, traffic management, and finance, there is enormous potential.

However, this also necessitates a strong focus on the further education and retraining of the workforce. While routine tasks may decline, the demand for skilled workers in areas such as data analysis, AI development, and expert knowledge for managing automated processes is growing. Governments, educational institutions, and businesses must therefore collaborate to ensure that this transformation is socially responsible.

Artificial General Intelligence (AGI)

Even though strong AI or Artificial General Intelligence (AGI) is still a thing of the future, predictions regularly emerge that don't rule out the emergence of this technology within the next few decades. AGI would be capable of learning independently, adapting to new contexts, and solving tasks with a similar range of abilities as humans. Whether, when, and how this will happen remains speculation. However, it is clear that such a development would have far-reaching consequences for the economy, politics, and society. Therefore, it makes sense to start thinking about ethical and regulatory guidelines today.

Related to this:

From language models to AGI (General Artificial Intelligence) – The ambitious goal behind “Stargate”

From technology to transformation: Why AI is more than a trend

The use of AI in companies is neither a short-term trend nor purely a matter of technology. Rather, it is a comprehensive transformation process that affects all levels of an organization – from the executive suite to operational staff. Companies face a multitude of challenges: The technological complexity requires a solid foundation of IT infrastructure and specific expertise. Data security and privacy place high demands on those responsible for managing sensitive information. Furthermore, the automation of processes raises liability issues, for example, if autonomous systems cause damage.

Change management plays a crucial role. Employees need to be made aware of the new opportunities and limitations of AI in order to reduce fears and reservations. Transparent processes, open communication, and targeted training programs are essential so that the workforce understands AI as an opportunity. If this succeeds, companies can benefit from significant productivity gains, reduce costs, and tap into new markets.

However, despite all the enthusiasm for the technological potential, it's crucial not to forget that AI also raises ethical questions. Risks of discrimination, lack of transparency, data protection, surveillance, and the danger of spreading misinformation are problems that can only be solved with clear guidelines and responsible action. Companies that successfully implement AI therefore rely on a balanced strategy comprising technological expertise, targeted data management, cultural change, and ethical awareness.

In the future, AI will continue to grow in importance, whether through multimodal applications, user-friendly platforms, or the increasing use of robotics and autonomous systems. This necessitates continuous education and training within society to close the skills gap and actively shape this transformation. It will also become increasingly crucial to establish legal and social frameworks that guarantee security, data protection, and fair competition.

Companies that recognize the strategic importance of AI early on can be among the winners of this technological transformation in the coming years. However, simply purchasing AI or launching a pilot project is not enough. Rather, a well-thought-out approach is needed that considers technical, personnel, organizational, and ethical aspects equally. If this succeeds, AI will become a powerful engine for innovation and value creation, not only generating new products and services but also offering the opportunity to sustainably transform the world of work and unlock human potential.

“If AI can be used for the benefit of humanity and societal risks can be addressed responsibly, it will be a true driver of growth and progress.” This perspective shows that AI is far more than a technical tool. It can become the epitome of a transformation that makes companies more agile and innovative, with effects extending to all areas of life. Companies should therefore not be deterred by the initial hurdles, but rather embark on the path to AI with courage, expertise, and a sense of responsibility.