From vision to realitySuccess factors in AI projects

The implementation of artificial intelligence (AI) promises companies considerable added value – from the automation of operational processes to the optimization of the value chain. However, despite the huge hype surrounding AI technologies, practice shows that more than twice as many AI projects fail compared to traditional digitalization initiatives. There are many reasons for this, ranging from an unclear strategic direction to insufficient data quality. Conventional project management methods often reach their limits with extensive AI projects.

In this article, we highlight the specific challenges that companies need to overcome when introducing customized AI solutions. Whether internally to increase efficiency or externally to create new customer experiences – the success of such AI projects depends largely on the right planning and implementation.

Die F&P Executive Solutions AG, eine führende Sozietät im Bereich Beratung und Interim Management, hat sich darauf spezialisiert, Unternehmen durch den gesamten Digitalisierungsprozess zu begleiten. Mit ihrer Expertise in der Implementierung von KI unterstützt F&P ihre Kunden dabei, Projekte erfolgreich umzusetzen und die Potenziale der Künstlichen Intelligenz voll auszuschöpfen.

Eight key success factors that are decisive for the successful implementation of AI projects are presented below.

Success starts with the use caseHow to choose the right approach

A successful start to an AI project begins with choosing the right use case. As Aristotle said: “A good start is half the success!” This is especially true for AI initiatives, as there is a wide range of possible applications for artificial intelligence – from optimizing operational processes to improving quality and developing new business models. However, the key challenge is to identify the right use case that meets both the company’s current challenges and the available resources.

What makes a use case “right”?

The right use case for AI projects is not just a question of creative ideas, but above all a strategic decision. Companies need to be clear about which pain points they want to address with AI and whether these are actually suitable for an AI-based solution. Some problems – such as those that require a high level of emotional intelligence or involve complex interrelationships – are less suitable for AI, as such systems often reach their limits in these areas.

Essential considerations when selecting the use case:

• Data quality: A solid database is the foundation of every AI project. Missing or inadequate data sources can significantly jeopardize success. Companies should therefore check whether the required data is available and of sufficient quality.
• Technical infrastructure: AI often requires computing and storage capacity that exceeds the existing IT infrastructure. This must be taken into account in project planning and budgeting in order to avoid technical obstacles at an early stage.
• Strategic alignment: The use case should be in line with the company’s overarching goals and be economically viable in the long term. A business case calculation helps to objectively evaluate different use cases.
• Feasibility: The financial, personnel and technological feasibility of the project must also be ensured. An AI project can only be successfully transferred to ongoing operations if all resources are correctly estimated and planned.

Methodical approach to identifying the use case

In order to identify the right use case for AI projects, a structured approach is crucial. It is advisable to use tools such as the “AI Navigator”, which helps to structure and prioritize possible use cases and analyse their benefits in detail. In addition, a “job-to-be-done” analysis helps to understand the specific needs of internal and external customers and to link these with value-adding AI solutions.

Through this methodical approach, F&P Executive Solutions ensures that the AI solutions developed are not only innovative, but also practical and value-enhancing. Ultimately, the success of an AI project is determined by how well the solution fits the specific needs of the company.

High-quality data as the basis for reliable AI results

“Data is the foundation of artificial intelligence – its quality determines success.”

The success of any AI project stands and falls with the quality and availability of the underlying data. In order to successfully train, test and use an AI in operations, not only large volumes of data are required, but also high data quality. This can come from a wide variety of sources – from internal business and operations support systems to production sensors and data from external partners. However, it is not enough for this data to simply be available; it must also fit the specific problem that the AI is supposed to solve and meet the highest quality standards.

Incorrect or insufficient data can distort the results of AI systems and therefore jeopardize the success of the entire project. This becomes particularly critical when an AI automatically evaluates data and makes decisions based on it. Incorrect data sets or distortions in the data lead to unreliable results that undermine confidence in the AI and its decision-making.

Quality assurance: not a product of chance

Systematic and methodical assurance of data quality is essential when implementing AI. Many companies are guided by recognized standards such as ISO 9000 (quality management) and ISO 27001 (information security). A comprehensive data management system must ensure that the data meets the highest standards in terms of availability, completeness, accuracy, confidentiality and recoverability.

Particular challenges in AI projects lie not only in identifying suitable data sources, but also in ensuring that the data collected is correct, complete, consistent and up-to-date. Typical distortions that must be avoided are:

• Time bias: Outdated data can lead to incorrect predictions, e.g. if an AI is supposed to optimize customer service but is accessing outdated data.
• Representative bias: When the training data does not cover all relevant user groups, such as an AI that filters job applications but is mainly based on data from male applicants.
• Distortions in data collection: If the data collected is incorrect or incomplete, for example in the case of an AI that is supposed to predict machine wear but has been trained on inaccurate data from defective sensors.

Successful AI projects start with a thorough understanding of data

Most of the work in an AI project – around 80% – is spent on understanding, preparing and analyzing the data. The reliability and effectiveness of AI can only be ensured through an intensive preparation phase. This involves several important steps:

• Data collection: From relevant, verified and validated sources.
• Data definition: To ensure that each piece of information is understood and correctly assigned.
• Data exploration: To recognize correlations and dependencies between the data and derive new insights from them.
• Data verification: To ensure that the data is complete, plausible and meaningful.

This structured process minimizes the risk of bias and ensures that the AI project is built on a solid database. Companies that carefully plan and implement this step lay the foundation for the success of their AI initiative.

Seamless AI integration for full potential development

Ein entscheidender Erfolgsfaktor bei KI-Projekten ist die nahtlose Integration des KI-Systems in die bestehende Unternehmensinfrastruktur. Eine KI-Lösung kann nur dann ihren vollen Nutzen entfalten, wenn sie effizient in die vorhandenen Systeme, Prozesse und Datenströme eingebettet ist. Ein häufiges Problem ist, dass Unternehmen den Aufwand, der für eine solche Integration erforderlich ist, unterschätzen. Oftmals wird eine KI-Lösung als isoliertes Tool implementiert, das nicht optimal mit anderen Unternehmenssystemen kommuniziert. Dies führt zu ineffizienten Prozessen und einer limitierten Nutzung der gewonnenen KI-Erkenntnisse.

The “out of the middle” approach

To avoid integration problems, integration should be planned as a central component of the AI project from the outset. A promising approach is the so-called “from the middle out” approach. Here, a pilot project is initially launched in a well-integrated area of the system landscape. This makes it possible to start the project on a small scale and expand it step by step – a procedure known as vertical scaling. This step-by-step approach allows quick successes to be achieved while gradually increasing the complexity of the project. It is important that the integration works smoothly at every stage.

Use of the findings beyond the original purpose

Another often underestimated aspect of AI projects is the multiple use of the knowledge and data generated by AI. The results of an AI system not only offer added value in their original area of application, but can often also be used profitably in other areas of the company. It is advisable to identify the potential for secondary use as early as the planning phase of an AI project. Creative brainstorming at the start of the project can often reveal unexpected additional benefits that should be integrated into the business case.

Added value through intelligent integration

The intelligent and well thought-out integration of AI systems not only creates additional application options for the technology within the company, but also promotes acceptance of the AI solution. The combination of a gradual introduction, broad utilization of the knowledge gained and complete system integration ensures that the AI solution creates sustainable added value and fits seamlessly into the company’s business processes.

Targeted know-how planning for smooth AI implementation

“Knowledge is power – and this is especially true in the world of artificial intelligence.”

An often underestimated success factor in AI projects is the need for specific expertise. Many companies tend to use previous IT or digitalization projects as a comparison without realizing that AI projects place completely different demands on expertise. For example, a programmer is not automatically a data expert, and a Power BI specialist cannot necessarily create the complex training data for a machine learning model.

The role of data experts

The data expert plays a key role in AI projects. But here, too, there are various specializations, such as data engineer, data analyst, data scientist or machine learning engineer. Choosing the right expert depends heavily on the specific requirements of the project. However, it is not enough to have only data experts in the team. Rather, a balanced mix of different functions and roles that are precisely coordinated with each other is required.

Making competencies transparent

To ensure that all the necessary skills are available in the project team, it is essential to create transparency about the existing skills in the company. There is often a lack of systematic recording and documentation of employees’ skills. A successful team is made up of different groups of experts: Business Experts, IT Experts, Management/Leadership Experts, Data Experts and Domain Experts. In some cases, it is even necessary to fill certain roles more than once in order to fully cover all the requirements of the AI project.

Skills mapping can help to identify suitable employees, uncover skills gaps and recognize capacity bottlenecks. This is the only way to address potential bottlenecks at an early stage and implement the project efficiently.

Strategic competence planning

Systematic planning of know-how requirements begins with a thorough analysis. The creation of a use case profile, which is carried out in five steps, is recommended here:

1. Goal definition: Clarification of the goal and the values to be created.
2. Identify challenges: Defining the specific problems that the project is intended to solve.
3. Solution hypothesis: Definition of the hypothesis as to how the problem can be solved.
4. Assumptions about tasks: Which basic tasks must be fulfilled?
5. Derivation of the required competencies: Creation of a competence board to document the necessary skills.

This structured approach not only ensures that existing capacities are used optimally, but also that potential skills gaps can be identified and rectified at an early stage. Only through careful and strategic skills planning can the best talents and skills be mobilized for an AI project.

Capacity planning and resource management

In addition to securing expertise, it is important to plan the available capacities of employees realistically. Project work that takes up less than 30% of a team member’s working time is often inefficient and should be avoided. Unsolvable bottlenecks and missing skills must be resolved before the project starts, whether by postponing the project, adjusting the scope or providing additional resources.

Strengthening trust in AIHow to create acceptance

“A system is only as good as it is accepted.”

One of the most important factors for the success of an AI project is the acceptance of the technology by its users. Even the most technically sophisticated system can fail if it meets with rejection. This risk is particularly high with AI projects, as the technology often arouses fears and reservations – be it concerns about job losses, fear of manipulation or loss of control. It is therefore essential to actively involve employees in the change process and win them over to working with AI.

Transparency as the key to trust

A frequently cited problem with AI implementation is the perception of the technology as a “black box” – a system whose decision-making process is opaque and non-transparent for the user. This lack of transparency leads to mistrust, especially in sensitive areas such as healthcare, human resources or the legal system. In order to gain the trust of users, transparency must be created from the outset. This means that the data used and how the algorithms work must be disclosed. Only then can users understand how and why an AI makes certain decisions.

User-friendliness must also be guaranteed. AI systems should integrate seamlessly into existing workflows and make work easier rather than more complicated. If users feel that they are supported by the technology, acceptance will increase.

Ethical principles and responsibility

The introduction of AI technologies must not be viewed in isolation. It has an impact on the outside world, society and the company’s customers. It is therefore crucial to adhere to ethical principles in the AI project. AI systems must respect people’s autonomy and privacy and must not have negative social consequences. Companies that openly communicate these ethical principles and integrate them into their AI strategies strengthen users’ trust in the technology.

Another important point is clear accountability for the decisions made by the AI. Humans must remain in control of the system. Regular review and approval of AI results by human experts can further promote trust in the technology.

Communication and involvement during development

In order to increase the acceptance of AI systems, communication about how the technology works should not just start at the implementation stage. Transparency should already be created during the development phase and users should be actively involved. This reduces uncertainty and skepticism and ensures that users develop a better understanding of the technology.

Another aspect is dealing openly with failures. Like any new technology, AI systems go through a learning curve. By disclosing which failures have occurred in the development process and how these have led to improvements, companies strengthen confidence in the continuous development and reliability of AI.

Acceptance through transparency and cooperation

The acceptance of AI systems is a decisive success factor. It can be promoted through transparency, clear communication and the involvement of users throughout the entire duration of the project. Only if users understand how AI works and what benefits it offers will they be willing to accept the technology and actively work with it.

Sustainable qualityWhy AI tests never end

“Quality begins with the intention set by management.” – Philip Crosby

Die Qualitätssicherung stellt in KI-Projekten eine besondere Herausforderung dar, da die Systeme kontinuierlich lernen und sich weiterentwickeln. Anders als bei traditionellen IT-Projekten endet die Qualitätssicherung nicht mit der Implementierung, sondern muss fortlaufend während des gesamten Lebenszyklus des KI-Systems sichergestellt werden. Viele Fehlerquellen, die aus klassischen Digitalisierungsprojekten bekannt sind – wie unklare Zieldefinitionen, mangelnde Planung oder unzureichendes Projektmanagement – beeinflussen auch die Qualität von KI-Projekten. Doch insbesondere die Datenqualität bleibt der zentrale Faktor.

Ongoing quality assurance in AI projects

A major difference to traditional IT projects is that quality assurance in AI projects not only takes place in the development and test phase, but must be monitored continuously during operation. AI systems change through constant learning, which makes continuous quality checks essential. Uncontrolled further development can lead to undesirable or unpredictable results.

Agile methods are particularly suitable for ensuring the necessary flexibility in project management. This allows adjustments to be implemented quickly and quality to be ensured on an ongoing basis. In addition to security, the traceability and transparency of AI systems must also be guaranteed. Systems should be developed “Secure by Design” and “Transparent and Traceable by Design” in order to make decisions traceable and promote trust in the results.

Extensive testing effort

The testing effort in AI projects significantly exceeds that of conventional IT projects. In addition to standard tests such as functional and integration tests, special AI tests are required. These include

• Adversarial testing: To check whether the AI model can be manipulated or deceived by targeted attacks.
• Fairness analyses: To identify and eliminate discriminatory patterns and distortions in the model.
• Ethics controls: To ensure that the system complies with ethical principles and has no negative social impact.

After the test phase, quality assurance should move on to permanent monitoring in order to continuously check the AI for potential sources of error and take immediate corrective action if necessary.

Effective change management and monitoring

Well-structured change management is crucial to ensure that all changes to the system are quality assured and fully supported by stakeholders. An AI oversight body can help monitor compliance with ethical guidelines and conduct regular quality audits. This body ensures that AI systems meet company requirements and that potential risks are identified at an early stage.

A/B tests for optimization

A/B tests are an effective tool for quality assurance in AI projects. In these tests, different versions of the AI model are run in parallel to directly compare their effectiveness. The variables here can be different algorithms, data sets or model configurations. The random assignment of the models to the test groups prevents distortions and ensures valid results. A/B tests provide in-depth insights into the behavior of the AI systems, which helps to select the most powerful and robust version.

Together to successBacking from the management

“Innovation distinguishes the leader from the follower.” – Steve Jobs

One of the key success factors for AI projects is the full support of management. This backing must not only be ensured at the start of AI implementation, but must also be maintained throughout the entire course of the project and during operations. Especially in AI projects, which are often associated with new technologies and far-reaching changes in work processes, management plays a central role. Both exaggerated scepticism and unrealistically high expectations can jeopardize the project, which is why balanced expectation management is necessary.

Uncertainty in management: building knowledge and trust

Management uncertainties often arise due to insufficient knowledge about AI technologies and their possible applications. Managers in particular, who pay close attention to information security, often have concerns as AI systems access and process extensive data sets. These uncertainties can significantly slow down or even prevent the progress of AI projects if they are not addressed in good time.

One important measure for overcoming these barriers is targeted management training. Workshops and information events can improve understanding of the risks and opportunities of AI. At the same time, the willingness to embrace change is strengthened if the benefits of the technology are presented clearly and comprehensibly.

Strategic expectation and risk management

AI projects often entail significant changes in the organization and work processes. To get the full support of management during AI implementation, it is crucial to practice professional project management. This includes comprehensive expectation management, regular progress reporting and effective risk management. These established structures create trust and ensure that management is informed about the project status and potential risks at all times.

A small, well-designed pilot project can help to demonstrate the potential of AI and achieve initial success without requiring large investments. Such pilot projects are often the basis for more extensive initiatives and show management the concrete added value of the technology. The measurement of success should be clearly defined and focused on measurable KPIs, such as cost savings, increased efficiency or employee and customer satisfaction.

Proactive involvement of management

For the long-term success of AI projects, it is important not only to use key management personnel as consultants, but also to actively involve them in planning and quality assurance. This involvement creates a sense of shared responsibility and improves transparency with regard to the AI methods and technologies used. Tools such as the Data-driven Business Model Canvas or the AI Navigator can be used to clearly illustrate the benefits and structure of AI projects.

Another important factor is measuring the adoption rate – it provides information on how quickly and effectively an AI system is accepted by end users in the company. A high adoption rate is often an indicator of the quality and relevance of the system, as it shows how well it meets the needs of users.

Management support as a key factor

Management backing is not only essential for the launch of an AI project, but also for its long-term success. Targeted involvement, transparent communication and comprehensive expectation management can reduce potential concerns and secure support. This forms the basis for successfully implementing AI projects and sustainably anchoring them in the company.

The most important success factorsWhat the survey reveals

A recent survey conducted by F&P Executive Solutions AG on LinkedIn made it clear that data quality and availability is considered the most important success factor in AI projects. 43% of respondents stated that insufficient data was the most serious problem in their projects. A further 21% cited the right know-how requirements as a decisive success factor, while 19% emphasized the choice of the right use case. The acceptance of AI solutions by employees was named as a key factor by 17% of respondents.

These findings confirm that careful planning and implementation of AI projects, especially in terms of securing high-quality data and providing the right expertise, is crucial. Companies must make targeted investments in these areas in order to bring their AI projects to a successful conclusion and benefit from them in the long term.

Are you ready to exploit the potential of AI?

Die Umfrage zeigt klar, welche Hürden in KI-Projekten überwunden werden müssen. Jetzt ist der richtige Zeitpunkt, um aktiv zu werden und die Möglichkeiten der Künstlichen Intelligenz voll auszuschöpfen. F&P Executive Solutions AG unterstützt Sie dabei mit einem maßgeschneiderten AI-Readyness-Assessment, um zu prüfen, wie gut Ihr Unternehmen für KI-Projekte aufgestellt ist. In einem unverbindlichen Beratungsgespräch analysieren wir gemeinsam Ihre spezifischen Herausforderungen und zeigen Ihnen, wie Sie die Erfolgsfaktoren in Ihrem Unternehmen optimal umsetzen können. Nutzen Sie die Chance, die Zukunft Ihrer Digitalisierungsstrategie zu gestalten – kontaktieren Sie uns noch heute.

Your contact for success factors in AI projects