1. Problem statement
In virtually every organization today, employees have direct access to powerful AI. At the same time, external consulting firms advise executive and supervisory boards with their own AI models, while major clients demand that supply chains and pricing logics be transparent and fairly algorithmized. The traditional knowledge and leadership pyramid is thus being impacted from three sides simultaneously: from below (employees), from the outside (consultants and clients), and from within (data scientists and VR). Leaders must renegotiate their credibility—not only with their teams, but also with consultants, supervisory boards, and the market. Transparency builds credibility; ethics and empathy create a new competitive advantage.
2. State of research - Key message
Topic Findings
_________________________________________________________________________________________________________________________________________________
AI as "Co-Manager" AI is already taking on operational management tasks (CEO Tang Yu, NetDragon, first robot possessing
an executive position).
Qustion of Trust A large proportion of employees and customers do not fully trust AI recommendations.
Transparency Becomes a Credibility Factor.
External Consultants Consulting firms like Deloitte, PwC, and KPMG deliver black box models; the board of directors and
management increasingly demand explainability reports.
Customer Transparency OEMs, automotive groups, and major IT customers require suppliers to disclose
algorithmic decision logic.
New Leadership Skills Four key areas: cognitive, digital, interpersonal, and self-leadership.
Ethics & Governance Managers are liable for algorithmic discrimination and data protection.
Employee-Empowerment AI-supported feedback systems give teams real-time insights into team dynamics and leadership quality.
__________________________________________________________________________________________________________________________________________________
3. New Challenges for Leaders
1. Real-Time Salience Testing
Employees and customers check every instruction against AI – unreflective "command culture" is immediately exposed.
2. Explainability of Decisions
AI provides data, but the interpretation is the responsibility of the leader. Those who fail to communicate results transparently
lose authority. External consultants deliver models whose codebase the client does not own.
3. Managing Hybrid Intelligence
Human intuition plus AI analysis require "bilingualism" in both emotion and algorithm.
4. Penta-Layer Moderation
Leaders must simultaneously synchronize employees (Layer 1), internal data scientists (Layer 2), external consultants (Layer 3),
customers (Layer 4), and VR (Layer 5).
5. Ethical Risks
Binding training data leads to unfair personal decisions; liability and reputation lie with the human.
6. Cultural Fragmentation
AI-savvy employees push for a leading role, while skeptics resist – division threatens if leadership
fails to provide an inclusive narrative.
________________________________________________________________________________________________________________________________________________
Component Specification
________________________________________________________________________________________________________________________________________________
1. Radical transparency documentation of all relevant AI models, data sources, filters, and decision logics is
accessible to employees, the Board of Directors, consultants, and clients.
2. Penta-Co-Creation workshops with employees, consultants, and clients to develop shared prompts and
governance rules.
3. Contractual AI clauses Consultants must provide source code, training data, filters, and explainability documentation.
Clients receive access to the SLA (Service Level Agreement) portal.
4. Double-Loop-Learning Leaders not only question results but also their own assumptions,
which lead to data selection – documented in the VR report.
5. Emotional Signal Strength Building trust through high empathy, as AI already covers the rationale.
6. Governance-Literacy Knowledge of the AI Act, ISO 42001, and EU data protection to protect employees and customers
and ensure compliance.
7. Agile Role Design Leadership roles are iteratively adapted via retrospectives – similar to agile software teams.
__________________________________________________________________________________________________________________________________________________
5. Action Framework for Companies
1. AI-Readiness Check
Self-audit: What is the level of AI competence at all management levels?
Suggested tool: Haufe Academy's AI Maturity Model
2. Leadership Lab
3-month experimental environments where managers and their teams and customers explore AI scenarios and
record reflection videos
3. Ethics Boards at Eye Level
Employee representatives, customer representatives, managers, and AI experts jointly decide on the use of appropriate models and filters
4. Prompt-to-Lead Trainings
Training in targeted prompting techniques to effectively use AI as a sparring partner for strategic questions
5. Transparent KPIs
The explanation rate (the proportion of decisions whose AI logic is publicly disclosed) becomes a key performance indicator for
management.
________________________________________________________________________________________________________________________________________________
6. Conclusion and Outlook
AI understands not only internal expert knowledge but also the traditional consultant monopoly. Credible leadership emerges when five voices are united simultaneously: employees, internal data scientists, external consultants, customers, and VR. This is because credibility is based on the ability to intertwine human values with algorithmic efficiency. In the long term, those organizations that understand shared leadership as a design task across all five levels—with transparency, ethics, and emotion as intangible competitive advantages - will prevail.
AI does not replace leadership or consultants—but leaders who use AI without a five-layer governance framework will be replaced by employees, customers, and the supervisory board together.
Transparency ends where the legitimate interests of the company begin. Managers therefore face a dilemma:
- Explainability internally and externally
(Employees, customers, board of directors, and consulting firms)
- Protection of confidentiality and competitive advantage
(Algorithms, data sources, prompts, filters, and business models)
Practical balancing rules that have proven effective in initial pilot projects:
1. "Traffic Light" Staged Model
- Green: Metadata, fairness scores, data origin (excluding raw data)
- Yellow: Aggregated model metrics, e.g., precision/recall – NDA required
- Red: Source code, feature lists, pre-processing – only for the internal AI ethics board and external auditors under confidentiality agreement
2. Synthetic Data and Model Cards
Instead of raw data, synthetic datasets and standardized model cards (fact sheets) are provided – sufficient
for plausibility checks, but without disclosing actual business figures.
3. Zero-Knowledge Proofs and Differential Privacy
In customer reverse audits, proof of fairness and non-discrimination can be provided through cryptographic certificates,
without disclosing the core algorithm.
4. Role-Based Transparency
- Employees gain insight into the sub-models relevant to them.
- Customers see only the segment model relevant to them.
- VR receives a complete fairness and risk report (red level under NDA).
- Consultants receive synthetic datasets and model cards (factsheets) and only data as required by the NDA.
5. Regulatory Safe Harbor Clauses
Contracts with consultants and customers stipulate that disclosures (e.g., through future EU implementing regulations)
are not considered breaches of contract, provided they are kept to a minimum.
Conclusion: Credible leadership in the AI age means using transparency as a "gradual regulator"—open enough for trust,
closed enough for sustainable secrecy and competitiveness.