Bio

(and why this is a crucial lesson for every AI adoption journey)

Once again, a major government IT project has been discontinued. Following earlier cancellations in the justice system and the police, an ambitious digitalisation programme in education has now also failed to deliver sufficient results to justify continued public investment.

By now, the pattern is familiar. What begins with high expectations and strategic ambition ends in disappointment, major revisions or complete termination.

The public response is predictable. Poor project management is blamed, as are the wrong suppliers, inadequate oversight by the commissioning authority or — often rather quickly — favouritism and conflicts of interest. Each of these explanations may play a role in individual cases. But when similar situations keep recurring across policy areas, it becomes increasingly difficult to dismiss them as isolated incidents.

If approximately one in five government IT projects fails to achieve its objectives or is terminated prematurely, there is probably something more structural at work. The problem is not primarily one of individuals or political parties, but of how we understand and organise digitalisation.

This observation becomes even more relevant when we broaden our perspective. The figures for AI projects — in both the public and private sectors — remain equally sobering. The exact percentages differ depending on the research report, but the overall message is remarkably consistent: as many as 70 to 80 per cent of AI projects never move beyond the pilot phase or fail to deliver their expected impact.

The parallel with traditional IT projects is too strong to be dismissed as coincidence.

What is really happening here? In this blog, I want not only to identify the underlying problem, but also to argue why a more integrated adoption journey is essential.

AI as a Magnifying Glass for an Old and Deeper Problem

What is striking about many failed government IT projects is that the technology itself is rarely the main problem. Systems are often delivered in technically correct form, data models are generally well designed and suppliers possess the necessary expertise. Yet the question remains: why do these projects ultimately deliver “too little”?

The answer usually lies not in what the system can do, but in how it is introduced into the organisation. Too often, insufficient attention is paid to the human factor.

IT projects are still frequently treated as neutral tools that make processes more efficient or organise information more effectively. In reality, they affect how decisions are made, who carries responsibility and how professionals understand their role.

This is particularly visible in sectors such as justice, policing and education. In these contexts, digitalisation is not only about efficiency. It also affects professional judgement, discretionary space and professional autonomy. When technology fails to respect or explicitly address these dimensions, a gap emerges between the system and the people expected to use it.

Artificial intelligence intensifies this tension. AI is not merely traditional IT that supports human activity. It structures decisions, formalises norms and makes previously implicit choices explicit. What was once weighed through professional consultation, experience and contextual judgement is translated into rules, models and scores.

Normative choices are therefore not only made; they are embedded and stabilised within systems.

It is hardly surprising, then, that AI projects often fail because of problems relating to trust and adoption. This is not necessarily because the model is technically incorrect, but because no one can clearly explain why the system produces particular outcomes, which values are embedded in it and when human intervention must remain decisive.

AI fails where meaning and responsibility have not been explicitly organised.

Implementation Is Not Adoption

A recurring misconception in both government IT and AI projects is that successful implementation will automatically lead to meaningful use and positive impact. There is a roadmap, a timeline, a procurement procedure and a go-live date. After that, the added value is expected to become self-evident.

Digitalisation — and AI in particular — does not work that way.

Technology changes not only processes, but also expectations, responsibilities and power relations. When these changes are not explicitly addressed and supported, the result is often not open resistance, but something more subtle: minimal compliance, circumvention or purely formal use without genuine integration into professional practice.

This is especially visible in education. Digital systems can function perfectly from an administrative perspective and still be experienced as a burden by teachers, school leaders or support staff. This is not necessarily because they oppose technology, but because the system does not sufficiently connect with their understanding of good education, professional practice or pedagogical quality.

This helps explain why so many projects ultimately “deliver too little”. It is not that they do nothing, but that they fail to establish a meaningful connection with the values and practices of their users.

Ethics as Value Alignment, Not an Afterthought

From this perspective, ethics takes on a very different role from the one it is often assigned.

Ethics is not an abstract framework that is applied at the end of a project, nor is it an obstacle to innovation. In successful adoption journeys, ethics functions as value alignment: the process of making explicit what different stakeholders consider important and determining how those values should guide the design, use and governance of technology.

When this conversation does not take place, values are simply assumed. They enter the system unnoticed through assumptions, data choices and design decisions, without anyone retaining meaningful control over them.

The result is technology that may be legally defensible but still creates moral and organisational friction.

Stakeholder involvement can play a crucial role here. It should not merely be used as a communication strategy to “create support”, but as a design space in which values and professional practices can become visible and open to discussion.

What do professionals need? What impact will the new system have? What might be lost, changed or strengthened?

By actively involving professionals, end users, citizens or students, organisations do not merely increase trust. They also develop better technology. Innovation becomes richer when it is informed by a diversity of perspectives.

From Failed Projects to an Integrated AI Adoption Journey

Many failed projects share a fragmented approach. Ethics, legal compliance, change management, technology and economics are treated as separate workstreams, each with its own timing, methods and logic.

As long as each element can be “checked off”, the project appears to be on track. In practice, however, the different parts fail to form a coherent whole.

A mature AI adoption journey requires a fundamentally different approach. It must be holistic rather than sequential. It should not focus on optimising one dimension in isolation, but on creating coherence between them.

Ethics supports value alignment and legitimacy. Legal expertise ensures compliance with relevant legislation. Change management guides shifts in professional roles and practices. Technology provides robust and explainable systems. Economics ensures sustainable long-term value creation.

Without this coherence, AI either remains stuck in pilot projects or results in solutions that are technically impressive but socially or organisationally unsustainable.

Conclusion: Coincidence Is No Longer an Adequate Explanation

The repeated termination of government IT projects is not simply a series of unfortunate coincidences. It is a structural signal that we continue to approach technology as though it were neutral, while in reality it has a profound normative impact.

The parallel with AI projects is therefore no coincidence either. What is visibly going wrong in government IT today foreshadows what will increasingly happen in companies tomorrow as AI becomes more deeply embedded in decision-making, governance and professional autonomy.

The central question is therefore not why this or that project was discontinued once again. The real question is why we continue to invest in technology without explicitly investing in value alignment, stakeholder dialogue and adoption.

That is where we stop launching yet another technology project and begin building a genuine AI adoption journey.

Key Considerations for Organisations

Start IT and AI initiatives not from the technology itself, but from explicit value alignment with stakeholders.

Use stakeholder deliberation as a design space to strengthen adoption, trust and innovation.

Develop ethics, legal compliance, change management, technology and economics together from the outset, rather than treating them as separate silos.

Do not treat AI as a project with a fixed end date, but as an ongoing learning process that requires monitoring and adjustment.

Measure success not only in terms of efficiency or cost reduction, but also through actual use, trust and perceived legitimacy.