AI in the SME Sector

This article sheds light on why it is not enough to start with pilot projects and why technology, organization and processes must be developed together for AI to have a lasting impact.

Most AI initiatives start with a specific use case: One area wants to forecast faster, another is experimenting with ChatGPT and a third is planning to use a co-pilot for internal knowledge searches. This is understandable and absolutely correct at first.

However, the operational reality often looks different:

• Data is stored in silos or is not qualitatively suitable for reliable modeling.
• Standard processes are either not available or not defined, which makes automation massively more difficult.
• Available tools and platforms are either too isolated or not prepared for AI workloads.

In addition, there is a particular dynamic in management: the view of AI is often characterized by wishful thinking, as tools such as ChatGPT suggest that everything "just works". However, IT sees the downsides: a lack of data integration, unclear operating models, compliance risks and technical complexity.

And this creates a dangerous gap between desire and reality.

An AI-enabled architecture is not abstract planning, but the prerequisite for scaling pilot projects, integrating systems and ensuring that use cases remain operable. It forms the basis for speed and governance at the same time and must enable both aspects: Experimentation on a small scale and reliability on a large scale.

In practice, many organizations are able to develop good solutions in specific areas, but without an overarching technical foundation, these remain isolated. A well thought-out AI architecture, on the other hand, acts as an amplifier: it helps to identify synergies between use cases, standardize data and assign responsibilities - and thus accelerates operational implementation.

Four interlinked dimensions have proven to be critical to success:

1. Data quality & Data architecture

   No AI success without clean, accessible and documented data. This applies not only to technical quality, but also to governance and role models (e.g. data owner, stewardship, auditability).

2. Scalable platform strategy

3. Roadmap sensitivity to hyperscalers

   Many AI functions will become standard features in platforms such as Microsoft, SAP or AWS in the coming months. Those who are aware of these developments can make smarter make-or-buy decisions - and avoid expensive in-house developments that will soon be obsolete.

4. Technology openness & integration

   KI must not be a tool focus. It is crucial that solutions can be integrated into existing system landscapes - for example through APIs, process integration or data layers.

Those who make technological decisions today are laying the foundations for the future. It's not about big architectural concepts, but about the ability to transform different initiatives into a growing whole.

In this role, IT must become the architect of scaling: It must enable the business departments to implement quickly without losing sight of long-term operability, maintainability or IT costs.

The decisive lever is to align technical and strategic realities at an early stage. This does not require 300-page concepts, but a clear roadmap, a shared vision and the ability to think about use cases systemically rather than in isolation.

Artificial intelligence is more than just a technology. It changes decision-making processes, responsibilities and expectations. It therefore presents organizations with perhaps the most challenging task of digitalization: changing behaviour, not just the system.

In many companies, there is currently a tension between the enthusiasm of the specialist departments and the uncertainty of the management. The specialist departments are developing their own ideas, while central teams are supposed to provide guidance - but often without the corresponding mandate.

This role - between enablement and governance - is really demanding. If you regulate too early, you put the brakes on innovation. If you coordinate too late, you lose synergies - and control.

In some companies, a familiar dilemma also arises: management expects quick success without first formulating a clear strategic direction. The operational reality then often looks like this: Specialist departments test, experiment, bring tools into play - and wait for backing that doesn't come. This so-called chicken-and-egg problem often leads to the transformation being wanted but not organized.

Cultural change does not need change campaigns. What it needs is:

• Space for participation.
• Employees who help shape the company.
• Departments that contribute their use cases.
• Teams that become visible and connectable within the organization.

If AI is not to be imposed from above - which hardly works in complex organizations - structures are needed that enable participation without allowing arbitrariness. The best results are achieved where a common framework for action is developed, but the professional proximity to practice remains.

Successful companies specifically establish interdisciplinary steering or strategy teams for this purpose. These do not see themselves as gatekeepers, but as orchestrators. They bring initiatives together, create transparency about ongoing projects, identify synergies and support initial pilots. This creates a cultural climate that focuses on coordination rather than control and on co-creation rather than silo thinking.

Culture is not an add-on, but the catalyst for trust. And trust is the prerequisite for any form of transformation.

Many companies do not start their AI initiatives with an official program, but with dedicated individuals or small teams from the areas of IT, analytics or automation. They take on responsibility before there is a formal mandate.

These initiatives are valuable, but they need structure. Because without orchestration there is a risk:

• Redundancies in the model and tool landscape
• Lack of traceability in decisions
• Unclear compliance situations for AI applications

At the same time, there is increasing pressure of expectation within the organization.

What is needed is a hybrid control approach:

• Enabling pilots - with technical and legal safeguards
• Document and evaluate use cases - standardized, but not formalistic
• Involve specialist departments, but keep them responsible.
• Establish central guard rails, but do not define everything in advance.

In this way, governance does not become a gatekeeper, but an enabler. It creates clarity about what is possible, sensible and responsible. Organizational governance is not an end in itself, but the framework in which innovation can unfold.

Nothing will ever be as slow as it is today - an insight that has become emblematic of technological development in recent years. The AI transformation in particular makes it clear how far two dynamics are diverging:

Platforms, models and functions are expanding rapidly. But while the technical curve is accelerating, decision-making processes, governance structures and role models are only changing gradually - often slowed down by a lack of resources, coordination requirements or uncertainty.

This discrepancy creates pressure. Departments want to get started, management demands strategic impetus, and IT or transformation are caught between desire and reality. The result: many companies are constantly accelerating, without a viable target image, without prioritization, without clarity as to where this movement should lead.

The task of those responsible for transformation is therefore not only to create structures, but also to enable the ability to act in motion:

• through smart goal setting and prioritization,
• through methodical clarity and transparency,
• and through a model that combines governance and pragmatism.

Many companies are currently experiencing the following: you don't need perfect plans - you need the courage to start with the right questions. The will to lead despite uncertainty. And the willingness not to manage transformation as a project, but to accompany it as a process.

In many conversations with CIOs, we hear a statement that sums it up: "Why should I invest strategically in something that is still so diffuse?"

This attitude is understandable, but also highly risky. Because in a technological environment that is developing exponentially, inaction is not a neutral decision, but a strategic choice with consequences.

What appears to be a hedge in the short term leads to a loss of control in the medium term:

• Departments create facts by independently implementing tools and services
• IT loses control over data, infrastructure and security standards
• Strategies do not emerge, but are replaced by operational actionism
• Distribution of budgets and responsibilities no longer follows the organization, but the momentum

As a result, CIOs and CDOs end up in the role of administrators instead of designers. If they do not invest because the target image is diffuse, they create exactly what should be avoided: Ambiguity, a lack of transparency and fragmentation.

But there is another way:

• A structured sparring process creates orientation without commitment
• Systematic prioritization of use cases brings impact without flying blind
• A proven methodological framework enables governance without inhibiting innovation

If you are faced with the question of how to take control in a fuzzy but dynamic environment, we invite you to start right there: With a workshop or a no-obligation sparring session. No pitch. Not a finished concept. But a concrete step towards clarity and impact.