Management Summary
Many AI initiatives fall short of their potential, even though the technology used is technically capable. In practice, the bottleneck rarely lies solely in the model, the tool, or the infrastructure. The key factor is whether AI is integrated into real-world business processes, responsibilities, data structures, governance rules, and competency profiles.
For IT and management, this sends a clear message: AI is not merely a technology project. AI is an organizational project with a technological core. Those who simply procure and roll out AI often end up with pilot projects, a proliferation of tools, and isolated efficiency gains. Those who, on the other hand, integrate AI with process design, data quality, leadership, change management, and skills development significantly increase the likelihood of measurable value creation.
Competency models can serve as a practical management tool in this context. They translate abstract requirements—such as AI readiness, AI governance, data literacy, critical evaluation of results, and responsible use—into concrete role profiles, learning paths, and leadership requirements. In doing so, they build a robust bridge between AI strategy and operational implementation.
Key Takeaway for Practice • AI success does not come from tools alone, but from a coordinated operating model that combines technology, processes, data, governance, and expertise. • IT, business units, management, and HR must work together to determine which skills are required for each AI application. • Competency models are not just an HR formality, but a management tool for managing scalability, risk, and value creation. |
1. Why AI Projects Often Fall Short in Practice
In many organizations, expectations for AI are high: faster processes, better decisions, lower costs, new services, and more productive employees. The operational reality is often more sobering. Individual teams are experimenting successfully, but the leap to stable, scalable, and measurable use does not happen automatically.
A typical pattern is technology-centricity. The first question is which tool to use, not which business problem to prioritize. As a result, there is often a lack of clear rationale for the benefits: Which process should be improved? Which decision should be supported? Which quality, cost, or time metric should change? Without this clarification, success remains open to interpretation, and scaling the solution becomes vulnerable to political criticism.
A second pattern is a lack of workflow integration. AI is treated as an additional tool alongside existing workflows. Employees then have to figure out for themselves when and how to use the system effectively. This leads to inconsistent use, shadow processes, and acceptance issues. However, value is only created when tasks, interfaces, approvals, checkpoints, and responsibilities are adapted.
A third pattern concerns data. Many companies have large data sets, but not necessarily data that is suitable, accessible, well-maintained, and legally usable for specific AI use cases. Data quality is therefore not a secondary consideration, but a top-priority implementation risk.
Finally, the issue of competence is underestimated. AI competence is too often reduced to prompting or tool operation. For productive use, however, organizations need a broader set of skills: a basic technical understanding, an understanding of data, an understanding of processes, the ability to critically evaluate results, ethical judgment, risk awareness, and communication skills.
2. Common Misconceptions and Better Management Logic
| Misconception | Operational risk | Improved Control Logic |
| "We need an AI tool." | Tool selection before defining the problem; low acceptance; unclear performance metrics. | Start with the use case: Define the problem, process, benefit hypothesis, and measurement criteria. |
| "IT implements AI." | Departments remain consumers rather than active participants; lack of process alignment. | Establish shared responsibility among IT, business, management, HR, data protection, and compliance. |
| "Training is enough." | One-time training sessions fizzle out; they aren't embedded in roles and routines. | Integrate competencies into role profiles, learning paths, leadership development, and performance logic. |
| "Governance slows down innovation." | Uncontrolled use, data privacy risks, bias, lack of traceability. | Understanding governance as scalability: clear guidelines enable safe use. |
| "The ROI is immediately apparent." | Decisions to terminate projects too early or painting a rosier picture of pilot projects. | Distinguish between learning benefits, process benefits, risk reduction, and financial benefits. |
3. What IT and Management Need to Manage Specifically
To ensure effective AI implementation, organizations should consistently address five key areas. These areas are not optional; together, they form the operational model for AI.
1. Business Problem and Value Proposition
Every AI use case requires a clear problem definition. Management and the business unit must define the expected value contribution: efficiency, quality, customer benefit, risk reduction, innovation capacity, or improved decision support. Without a clear vision, AI becomes a demonstration of technical capabilities rather than a lever for creating value.
2. Process and Workflow Integration
AI must be integrated into existing or newly designed workflows. This includes input points, verification steps, decision-making authority, escalation procedures, and documentation. It is particularly important to determine when humans make decisions, when AI provides support, and when automation is permissible.
3. Data Quality and Data Governance
Data must be available, accurate, well-maintained, relevant, and legally usable. This includes responsibilities for data maintenance, access rights, data protection, data provenance, and quality assurance. Without data governance, AI becomes either unreliable or unscalable.
4. AI Governance and Risk Management
Organizations need clear rules governing permitted and prohibited uses of AI. These include approval processes, human oversight, documentation, handling of sensitive data, bias assessment, security requirements, and accountability. Governance is not a bureaucratic exercise, but rather a matter of risk management and building trust.
5. Competency Development and Change Management
Employees and managers must understand what AI is capable of, where its limitations lie, and how to critically evaluate its results. At the same time, concerns, changes in roles, and issues of acceptance must be actively addressed. Communication is not merely a side note; it is an integral part of the implementation process.
4. Competency Models as a Practical Bridge Between Strategy and Implementation
A competency model describes the skills, knowledge, attitudes, and behaviors required to perform effectively in specific roles. In the context of AI, such a model must go beyond traditional digital skills. What matters is not only whether employees can use a tool, but whether they can apply AI in a meaningful, critical, legally compliant, and value-driven manner.
Managers face different priorities than technical specialists or business users. Managers must strategically position AI, set priorities, understand risks, clarify responsibilities, and communicate change credibly. IT and data teams need deeper knowledge of architecture, data quality, security, model limitations, and integration. Business units must be able to identify use cases, validate results from a business perspective, and take responsibility for decisions.
This turns the competency model into a translation tool. It highlights which roles require which competencies, what development measures are needed, and where risks arise due to a lack of skills. This is relevant for HR because it allows recruiting, training, leadership development, performance management, and succession planning to be aligned with the AI strategy.
Four Competency Dimensions for AI Implementation • Technical expertise: Basic understanding of AI, data, system limitations, security, and integration. • Strategic expertise: use case prioritization, value contribution, scaling logic, and decision-making ability. • Communication skills: transparent communication, engagement, building acceptance, and cross-functional collaboration. • Ethical and governance-related competence: data protection, fairness, transparency, human oversight, and responsible use. |
5. Practical Roadmap for Organizations
- Step 1: Prioritize AI Use Cases
Don't start with technology; start with business problems. Evaluate use cases based on benefits, feasibility, data availability, risk, and strategic relevance.
- Step 2: Clarify Roles and Responsibilities
Determine who is responsible for the subject matter, who oversees the technical implementation, who assesses risks, and who approves decisions.
- Step 3: Identify the competencies required for each role
Define the minimum competencies required for leadership, IT, functional departments, HR, compliance, and data protection.
- Step 4: Create learning paths
Combine general AI literacy for all relevant users with role-specific in-depth training. Training must be based on real-world use cases.
- Step 5: Put Governance into Practice
Design guidelines, approvals, documentation requirements, escalation procedures, and control mechanisms so that they can be effectively used in day-to-day work.
- Step 6: Measure and Scale the Impact
Evaluate pilot projects based on defined metrics. Do not scale up until the benefits, risks, data, and organizational requirements are well-established.
6. Checklist for Decision-Makers
- Has a specific business problem been defined for every AI use case?
- Are the benefit hypothesis and key performance indicators defined before the project begins?
- Has it been determined how AI will be integrated into the target process?
- Have data quality, data access, and data protection been adequately assessed?
- Are there clear responsibilities for approval, use, monitoring, and escalation?
- Are executives, IT, line departments, and HR integrated into a common operating model?
- Are AI competencies incorporated into job descriptions, learning paths, and leadership development?
- Is there a communication strategy for acceptance, transparency, and managing expectations?
- Are risks such as errors, bias, lack of transparency, and shadow usage actively managed?
- Is a clear distinction made between the pilot phase, production operation, and scaling?
Conclusion
The key management message is this: AI does not succeed simply by implementing a powerful system. AI succeeds when organizations embed it into their processes, responsibilities, governance, and skills development.
For IT and management, this means a clear division of labor. IT creates a secure, integrable, and scalable technical foundation. Management prioritizes value contributions, makes strategic decisions, and is responsible for the operating model. HR and leaders ensure that the necessary competencies are not developed by chance, but are systematically built up.
In this context, competency models are not merely an additional academic topic. They are a practical tool for linking AI strategy, role requirements, professional development, governance, and implementation. Organizations that consistently establish this connection will manage AI more effectively, better control risks, and generate value in a more resilient manner.
Short excerpts from the original article
The abstract summarizes the main points of the original article and refers in particular to works on digital transformation, AI readiness, AI implementation, AI governance, and competency models, including Vial (2019), Verhoef et al. (2021), Jöhnk et al. (2021), Raisch & Krakowski (2021), Lee et al. (2023), NIST (2024), Ryseff et al. (2024), the World Economic Forum (2025), and Onnen (2024, 2025a, 2025b, 2026).
