The paradox is obvious: everyone is talking about AI as the solution to demographic change. Yet most projects fail to deliver. Why is that?
The real problem isn’t the technology
When AI projects fail, fingers are quickly pointed at the technology: too immature, too expensive, too complex. But this explanation falls short. The technology works – what doesn’t work is the foundation on which it is deployed.
Based on our experience from dozens of projects with German SMEs, the uncomfortable truth is this: this is not a debate about technology. It is a knowledge crisis within the German economy.
Companies do not fail because of AI – they fail because their own knowledge is invisible, scattered and tied to individual minds. AI can only work if it has access to a solid knowledge base. And it is precisely this foundation that is missing in most organisations.
Demographic pressures are exacerbating the problem
As workforces age, decades of specialist knowledge are quietly disappearing from companies. This is increasingly hampering companies’ effectiveness.
Imagine the following scenario: what would happen in your company if three key personnel were to retire tomorrow? The production manager who has known the plant for 25 years. The sales manager who has all the key customer relationships memorised. The IT administrator who knows why that one legacy interface is configured the way it is.
This is not a hypothetical question. In many organisations, it is a matter of months, not years.
The crucial follow-up question: Do you really know where your critical specialist knowledge lies – or do you just know who possesses it? The difference is existential. If you only know who possesses the knowledge, then you face a personnel risk. If that person leaves, the knowledge goes with them.
Data Maturity: The forgotten prerequisite
If an AI project were to start tomorrow – could you explain exactly which systems, data and rules it is actually permitted to use?
Most companies cannot. Not because of negligence, but because IT landscapes that have evolved over time have become so complex that nobody has a complete overview any more.
This is precisely the main reason why AI projects fail: it is not the technology that is the problem – it is the maturity of data availability. What we call ‘Data Maturity’. If this foundation is missing, even the best AI cannot achieve anything.
Three pillars for successful AI projects
How can this be changed? Based on our project experience, three success factors have emerged:
1. Making structures visible
The principle sounds trivial: we can only add value if we know how value is actually created within the organisation. Yet in practice, this value creation logic is usually implicit – it lies in informal processes, in habits, and in the minds of individual staff members.
A medium-sized industrial company wanted to scale its service quality globally. To achieve this, we first had to understand: Which departments are involved? What knowledge do they need? Where is this knowledge currently located – in software systems, in documents, or only in the minds of individual experts?
The findings were revealing: knowledge was fragmented, spread across systems, documents and people’s minds. Value was created solely through the interplay of these contributions. And the organisation’s actual logic was not documented anywhere – meaning it was neither scalable nor automatable.
2. Making data available
The second principle: it is not people who need to scale – but the knowledge they possess.
In practical terms, this means: expert knowledge must be made accessible to all staff, regardless of department or location. Data silos must be broken down so that AI has a foundation to build upon. And knowledge must be accessible round the clock – even when the expert is not currently available.
In view of demographic change, knowledge can no longer be a personal attribute. It must become an organisational resource.
3. Using automation correctly
Only once the first two pillars are in place does AI automation make sense. The correct approach follows a clear sequence: first, identify the process where time is currently being wasted; then formalise the required knowledge; and only then instruct the AI, giving it access to relevant data and clear rules.
The aim here is for AI to do the groundwork, whilst humans make the decisions. It is about reducing the workload, not replacing people.
What is possible when the foundations are right
A real-world example illustrates the potential: we automated route safety checks for a major transmission system operator. Hazards along power lines must be identified regularly – growing trees, new developments, changes to the road network. This used to be a manual process that took months.
The AI solution automatically detects and classifies these hazards according to 49 land-use categories. The result: a 97 per cent reduction in workload. Months of manual work were reduced to minutes.
But – and this is crucial – the company has not replaced any people. It has reclaimed time. Time for thinking, learning and making better decisions. The final assessment remains with humans.
Conclusion: AI is insurance against instability, not the risk
Our three key insights are:
Knowledge must be systematised. As long as critical specialist knowledge remains locked inside individual minds, it is a risk – not an asset.
Data maturity is the key prerequisite. It is not the technology that determines success or failure – but the maturity of your data landscape.
AI is not a risk – but the greatest safeguard for stability we have. Used correctly, AI makes companies less dependent on individuals and more resilient to demographic change.
The future of the German economy is not determined by models or algorithms. It hinges on the question: Who owns the knowledge – individual minds or the organisation?
This article is based on the webcast ‘Using data strategically – creating more value through AI’ by the Materna SME Initiative. For more information, visit materna.de.
