AI in Controlling, AI in HR 2025: Between Progress, Loss of Control, and Ethical Dead Ends. Made with Flux1.ai by Gary Owl

Gary Owl 5 months ago2025-10-13

AI in Controlling and HR 2025: Between Progress, Loss of Control, and Ethical Dead Ends

A Critical Analysis: The Great AI Transformation and Its Blind Spots

In 2025, AI is driving innovation in the DACH region, promising efficiency and better decisions. However, are data protection violations and algorithmic discrimination undermining progress? This article examines the challenges of AI implementation in controlling and HR.

This article, based on up-to-date sources (2024/2025), highlights the challenges and opportunities of AI use, analyzes the impact of the EU AI Act and ISO/IEC 42001, and provides practical recommendations for companies and policymakers.

By Gary Owl | June 01, 2025 | AI, Controlling, HR, Illusion of Steering

2. Methodological Approach

This article is based on current studies, market reports, and regulatory documents from 2024/2025, as well as case studies and expert interviews from the DACH region.

3. Theoretical Framework and Methodology: Interdisciplinary Perspectives, Definitions, and Methodological Approaches to AI Integration in Human Resources

AI systems are increasingly permeating strategic and operational HR processes, with their implementation posing complex interdisciplinary challenges. Recent studies indicate that 73% of companies will use AI-supported tools in recruiting by 2025, while 68% of HR managers express ethical concerns about algorithmic decision-making. This paradox underscores the need for a theoretical framework integrating technological, legal, and psychological dimensions.

3.1. Interdisciplinary Perspectives: From Ethics to Organizational Psychology

Furthermore, understanding AI in controlling is essential for HR managers to navigate these challenges effectively.

3.1.1. Ethical-Legal Governance Structures

The AI Best Practice Guidelines of the MacEwan University (September 2024) provides a practical framework mandating transparency, data minimization, and human decision-making authority. These guidelines align with the EU AI Act, which has banned emotional inference systems in workplaces since August 2024 and subjects high-risk AI applications in personnel management to strict conformity assessments. Notably, an analysis of 120 companies reveals only 22% fully implement the required documentation of AI decision processes, highlighting implementation gaps.

3.1.2. Psychological Implications of Algorithmic HR Decisions

Recent occupational psychology studies demonstrate that AI-supported assessment systems induce 34% higher stress levels in candidates compared to traditional methods, particularly when decision logic remains opaque. Conversely, adaptive learning systems like IBM Watson Talent enable 40% faster competency development through personalized training recommendations. This dichotomy necessitates integrating behavioral economic models to overcome acceptance barriers and compensate for cognitive biases in training data.

3.1.3. Economic Efficiency vs. Human-Centered Work Design

While AI-powered screening can reduce recruitment costs by up to 58%, a longitudinal study by ZHAW shows automated hiring decisions increase turnover rates within the first 12 months by 19%. This underscores the need for hybrid decision models, as proposed by the HR Tech Ethics Board: AI serves as a predictive support system, while final personnel decisions remain with trained HR experts.

3.2. Definition and Characteristics of AI Systems in HR Management

3.2.1. Structural Differentiation of AI Technologies

Unlike classical HR IT systems, AI applications are defined by four core characteristics:

Context-Adaptive Learning: Systems like LinkedIn Talent Insights use reinforcement learning to dynamically adjust candidate profiles to changing job requirements.
Predictive Analytics: People analytics tools forecast turnover risks with 89% accuracy by correlating 120+ variables from HRIS data.
Natural Language Processing: Chatbots like Mya automate 73% of initial candidate communication, though emotional intelligence simulations remain legally contentious.
Autonomous Decision Architectures: Fully automated onboarding systems optimize training plans in real-time but risk overfitting to historical data.

3.2.2. Regulatory Classification and Compliance Requirements

The EU AI Act categorizes AI systems into four risk classes, with HR management tools uniformly classified as high-risk under Annex III. By 2027, HR departments must:

Document all training datasets and decision protocols
Conduct annual audits through independent certification bodies
Implement human-in-the-loop controls for critical decisions
ISO/IEC 42001 complements these requirements with technical specifications for bias reduction, including synthetic data augmentation and continuous drift monitoring.

3.3. Methodological Approaches: Empirical Evaluation and Design-Oriented Research

3.3.1. Mixed-Methods Approaches for Impact Analysis

A meta-analysis of 47 studies reveals triangulative research designs best capture AI integration complexity:

Quantitative: Longitudinal performance metrics (e.g., time-to-hire reduction vs. quality-of-hire improvement)
Qualitative: In-depth interviews mapping employee acceptance patterns
Experimental: A/B testing of algorithm versions in live environments

3.3.2. People Analytics as Strategic Management Tool

Modern HR departments deploy predictive analytics across three core areas:

Talent Acquisition: Neural networks identify passive candidates through 500+ social media signals per profile.
Workforce Planning: Monte Carlo simulations model staffing needs considering macroeconomic indicators.
Employee Experience: Sentiment analysis correlates Slack communication patterns with engagement scores (r=0.82).

3.3.3. Design Principles for Human-Centric AI Systems

Derived from HR Tech Ethics Board guidelines, seven design principles emerge:

Ethical Fallback: Deactivation protocols upon detecting discriminatory patterns.
Transparent Explainability: AI recommendations require interpretable feature importance scores.
Controlled Autonomy: Critical decisions need dual HR expert validation.
Dynamic Adaptability: Monthly concept drift checks and model retraining.
Data Parity: Training datasets must represent ≥30% underrepresented groups.
Privacy Protection: Differential privacy ensures anonymity in subgroup analyses.
Continuous Learning: Feedback loops integrate employee evaluations into model optimization.

4. Controlling in 2025: Excel Graveyards and the Illusion of Steering

4.1. The Status Quo: More Technology, But Not More Steering

Despite available BI and AI tools, 41% of surveyed controllers in German large enterprises spend more than 60% of their working time on manual data aggregation. Robot-supported process automation and predictive analytics using artificial intelligence are still just dreams of the future in many places. Excel remains the dominant tool, even though real-time data analysis and automated reporting are technically possible.

4.2. The Controller of the Future: From Number Cruncher to AI Manager

The role of the controller is fundamentally changing. AI controllers monitor algorithms, calibrate models, and translate results into strategic recommendations (Becker et al., 2024).

5. HR and Recruiting: Between Efficiency, Data Protection, and Discrimination

5.1. AI in Recruiting: Blessing and Curse

AI reduces hiring costs by up to 40% among the analyzed DAX companies (Bersin, 2025), but in practice, 62% of the AI algorithms analyzed by ETH Zurich in Swiss and German companies prioritize younger applicants under 35, regardless of their qualifications (ETH Zurich, 2025).

5.2. Data Protection: Applicant Data as Training Material?

The EU AI Act and the GDPR prohibit the use of personal data without consent. Nevertheless, 22% of surveyed HR departments in the DACH region use applicant data for AI training (Rexx Systems, 2024). In Switzerland, a much-discussed data leak occurred in 2024: an AI system disclosed the home address of an applicant upon targeted request. The Swiss data protection authority subsequently demanded stricter controls and audits for HR AI applications.

5.3. Discrimination and Subjectivity: Old Problems in a New Guise

Skills-based hiring increases diversity by 28% of the new hires analyzed but is only used by 34% of surveyed HR departments in the DACH region (HRXConnect, 2025).

6. Ethical and Organizational Challenges: What Companies Need to Do Now

6.1. Governance and Responsibility

Only 12% of companies in the DACH region already meet the requirements of ISO/IEC 42001 for AI risk management (Emerald Insight, 2025). The EU AI Act requires clear responsibilities and provides for fines of up to €35 million or 7% of global turnover for violations (Clarius Legal, 2025).

6.2. Human and Machine: The Right Balance

AI should not replace people but empower them. The best results are achieved when AI and humans work in tandem-with clear roles, responsibilities, and feedback loops.

6.3. Leadership and Cultural Change

The biggest obstacle is not technology but leadership culture. Bold, visionary leadership is needed that not only allows change but actively shapes it.

7. Recommendations for Action: Ways Out of the Dead End

For Companies:
– Introduction of a comprehensive AI governance framework according to ISO/IEC 42001
– Certification programs and further training for AI skills in controlling and HR
– Technical and organizational measures to protect sensitive data

For Policymakers and Regulators:
– Stricter regulation and implementation of the EU AI Act
– Promotion of transparency and traceability in automated decision-making processes
– Expansion of information and objection rights for applicants

8. Conclusion: Turning Everything Upside Down-But With Reason

The AI revolution is a reality. It is radically transforming controlling, HR, and many other areas. But efficiency gains and automation are not an end in themselves. Without ethical guidelines, transparency, and a real redefinition of roles and responsibilities, there is a risk of loss of control, loss of trust, and a degradation of human work. Companies that prioritize AI governance, skills-based hiring, and ISO/IEC 42001 will secure long-term competitive advantages. Policymakers must consistently implement the EU AI Act-especially in the DACH region.