We argue that AI amplifies existing patterns in power and decision-making.
We insist governance and human judgement come first.
We call for traceable, contestable systems that protect dignity.
AI (artificial intelligence) amplifies what already exists. That’s the short version. The longer, more important version: AI holds up a mirror to how we decide, collaborate and account for our actions. If you mainly see software, you’ll miss the real intervention: in roles and mandates, in assumptions and yardsticks, in legitimacy and remediation.
We tend to approach AI as an implementation. But implementing without re‑examination is accelerating with the handbrake on. The core question is not “does the model work?”, but “which patterns does this system reinforce—and who bears the consequences?”.
Relationship before technology. Systems shape culture and culture shapes systems. If a model determines who receives attention, we are steering relationships; if a score determines who gets rejected, we are steering dignity. That’s why neutrality without governance is an illusion. Every model is a design choice with assumptions, data and thresholds. Secure revocability and contestability—or accept that efficiency will mask moral judgement.
Learning is double. AI excels at single‑loop learning: getting better at what we already do. Organisations only flourish when the question and the success criteria change as well (double‑loop). Otherwise we end up with beautiful dashboards and precisely the wrong decisions. The remedy is as prosaic as it is powerful: build in structured reflection on outcomes, organise dialogical decision‑making and cultivate judgement as a team capability.
Legitimacy is evidenced action. Carefulness only counts if it is traceable. Document purpose, data, bounds, performance and exceptions. Make logbooks, model cards and risk files tools—not appendices. What you automate, you must also be able to explain, contest and—if needed—repair.
Governing is the ability to hold tension. AI seduces us toward centralisation, standardisation and speed. Yet quality emerges where we refuse to look away from tension: between efficiency and resilience, between automation and human dignity, between openness and security. Governance here means choosing with conditions: where do we standardise and where do we let variety live? Where do we delegate to systems and where does human judgement remain at the line?
Why now? The European framework (General Data Protection Regulation, Digital Services Act, Data Act, NIS2 and the forthcoming application regime of the EU AI Act) makes careful design not only morally right, but also a managerial and legal necessity. Those who design with counter‑force today will gain in speed, trust and continuity tomorrow.
Read my full article: here