AI Center of Excellence: How Swiss SMEs Coordinate Their AI Transformation

The organizational gap: Why isolated AI tools don't scale

Swiss companies invest more heavily in AI than the European average, yet results fall short of expectations. The reason is rarely technical. McKinsey documents in the State of Organizations 2026: 83% of surveyed companies recognize the need for a new operating model to scale AI—yet 65% have no coordination structure. Deloitte puts it plainly: the leap is from isolated AI tools to an enterprise-wide AI operating model. Without an organizational answer, AI remains stuck in the pilot trap.

This gap is particularly pronounced in Swiss SMEs. With typically 10 to 250 employees, they lack the critical mass for a dedicated AI department. At the same time, regulatory requirements are rising through revDSG and the extraterritorial effect of the EU AI Act. What's missing is a lean, pragmatic coordination unit: the AI Center of Excellence.

What an AI Center of Excellence is—and what it isn't

An AI CoE is not a technology department. It is a coordination function that bundles strategy, governance, enablement and monitoring. The goal is not to centrally steer all AI projects, but to set standards, share knowledge and enable decentralized execution. The CoE answers questions like: Which use cases do we prioritize? What data may we use and how? How do we measure success? How do we enable business units?

McKinsey documents that companies with a functioning CoE achieve a 2.3 times higher success rate in scaling AI initiatives. The difference lies in the ability to consolidate learnings, avoid redundant investments and address governance risks early.

Start small: The single-person CoE

Ralf Blaschke from Accenture formulates a decisive insight in the Swiss AI Podcast: 'A Center of Excellence can be a single person deeply embedded in AI expertise.' For many Swiss SMEs, this is the most realistic starting point. This person takes on four core tasks: mapping existing AI initiatives, defining evaluation criteria for new use cases, establishing initial governance guardrails and building an internal knowledge network.

The typical profile: someone with technical understanding, business acumen and access to executive leadership. Often they come from the innovation or digitalization function. The decisive factor is not the title but the mandate: this person must be empowered to coordinate cross-functionally and have direct access to strategic decisions.

Scaling to a cross-functional team

When multiple parallel AI initiatives run or the company reaches a larger size, the single-person model reaches its limits. The CoE then grows into a small, cross-functional team. BCG recommends in their four-phase model institutionalizing governance and coordination in months 10 to 15 of AI scaling. For Swiss SMEs, this typically means a small core team with clear roles.

• AI Strategist: prioritization, roadmap, business case validation
• Data Steward: data quality, access, architecture
• Compliance Officer: revDSG, EU AI Act, internal guidelines
• Change Manager: training, communication, adoption measurement

These roles do not necessarily have to be full-time positions. In many SMEs, existing managers take on these functions part-time—the decisive factor is formal mandating and regular coordination. The World Economic Forum identifies a new role in 2026: the 'AI workforce manager' who coordinates human-machine hybrid teams. In practice, this function is often assumed by the Change Manager in the CoE.

The four core functions of an AI CoE

1. Strategic steering

The CoE curates the AI roadmap. It evaluates use cases according to business value, feasibility and strategic fit. It orchestrates resource allocation and ensures AI initiatives contribute to corporate objectives. Typical artifact: a quarterly updated priority list with clear go/no-go criteria.

2. Governance and compliance

The CoE defines, documents and monitors guardrails. These include data protection standards, bias checks, transparency requirements and escalation paths. In Switzerland, this area is not optional: 82% of Swiss companies have weak to medium data ecosystems according to the Swiss AI Observatory 2026—governance closes this structural gap.

3. Enablement and capability building

The CoE builds internal competence. It organizes training, curates best practices and provides reusable building blocks (prompts, workflows, datasets). It addresses the 76% paradox: most employees remain AI novices because structured enablement is missing.

4. Monitoring and ROI measurement

The CoE establishes a consistent measurement framework. It tracks not only technical KPIs (model accuracy, latency) but business metrics: time saved, error reduction, revenue effects. Without unified metrics, the business case remains diffuse and scaling stalls.

Case study: A Swiss industrial supplier builds its CoE

A Zurich-based supplier starts in 2025 with several isolated AI pilots: predictive maintenance in production, automated quotation generation in sales and quality control via computer vision. Each pilot runs in its department without exchange. Executive leadership appoints the Head of Digitalization as AI coordinator in Q3—the single-person CoE is born.

First measure: use-case register. The coordinator maps all initiatives, identifies data overlaps and discovers that two departments are licensing the same external API. Second measure: governance checklist. Together with legal counsel, she develops a one-page checklist for new use cases—data protection, bias, explainability. Third measure: monthly steering committee with executive leadership and business unit heads. After a few months, redundant costs were noticeably reduced, a unified ROI framework was established and new use cases were prioritized in the pipeline.

In the second year, the CoE grows: the coordinator becomes a full-time AI strategist, the IT lead assumes the data steward role part-time, the HR lead coordinates change and training. The cross-functional trio meets weekly, the steering committee remains monthly. Result: multiple productive AI applications, measurable efficiency gains and a workforce that understands AI as a tool, not a threat.

Three common pitfalls—and how to avoid them

Pitfall 1: The CoE becomes a bottleneck

If every AI initiative must be approved by the CoE, a bottleneck emerges. Solution: the CoE defines guardrails and standards but does not decide on every use case. Business units remain accountable—the CoE enables and reviews ex post.

Pitfall 2: Too theoretical, not operational enough

A CoE that produces strategy papers but is never involved in real projects loses credibility. Solution: at least one person in the CoE actively participates in pilot projects. This keeps the CoE grounded and learning from practice.

Pitfall 3: Missing connection to executive leadership

Without direct access to the C-suite, the CoE remains toothless. Solution: the CoE lead reports directly to the CEO or COO, the monthly steering committee is C-level staffed and use-case prioritization is an executive decision.

When to scale, when to outsource?

Not every SME must build a full internal CoE. The decision depends on three factors: number of parallel AI initiatives, strategic importance of AI and internal capacity. As a rule of thumb: with only a small number of active use cases and limited internal AI expertise, a hybrid model is often more efficient—an internal coordinator plus external sparring partner for strategy, governance and enablement.

KI-Outsourcing.ch operates precisely in this hybrid space: we act as an external AI division for Swiss SMEs that want to scale AI without building a complete internal structure. The internal coordinator remains the strategic point of contact, we deliver governance frameworks, ROI measurement, pilot support and change enablement. Once the organization is mature enough, it gradually internalizes—or remains in hybrid mode if that is economically more sensible.

Checklist: The first 90 days of your AI CoE

1. Clarify mandate: Who decides? Who reports where? What budget is available?
2. Create use-case inventory: Map all running and planned AI initiatives, name owners, document status.
3. Define governance fundamentals: One-page checklist for data protection, bias, transparency—pragmatic, not academic.
4. Establish steering committee: Monthly format with executive leadership and business unit heads, fixed agenda (priorities, risks, learnings).
5. Identify quick wins: One to two use cases that deliver value in 60 days—visibility creates momentum.
6. Sketch enablement plan: Who needs what training when? How do we share learnings? Which best practices do we document?

These six steps create the foundation. The CoE is not a big-bang project but an iterative build. Start lean, learn fast, scale as needed.

Conclusion: Organizational maturity determines AI ROI

The technological building blocks for AI scaling are available. The decisive variable is organizational. McKinsey shows: 83% recognize the need for a new operating model, yet only a few implement it systematically. An AI Center of Excellence is the practical answer—not a monstrous structure but a lean coordination function that can start with a single person in Swiss SMEs.

The investment is manageable, the leverage enormous. Companies with a functioning CoE scale AI 2.3 times more successfully. They avoid redundant costs, reduce governance risks and enable their workforce systematically. The first step is a mandate decision: Who coordinates? Starting when? With what authority? Technological transformation follows organizational—not the other way around.