Own GPT for the Company: Is it Worthwhile for Swiss SMEs?

Nearly half of Swiss SMEs (45%) now see Artificial Intelligence as a clear advantage for their business operations. This is according to a recent survey by kmu.admin.ch from 2025. This figure is encouraging and reflects the growing awareness that AI is not just a hype, but an essential tool for the future.

However, there is often a gap between recognising an advantage and its secure, efficient implementation. Many Swiss SMEs face the challenge of implementing AI solutions without losing control over their sensitive company data. Especially in a country with strict data protection laws like Switzerland, the question of data sovereignty is crucial. This raises the question: is a proprietary GPT system the answer for greater security and tailored efficiency?

A blanket answer of "Yes, absolutely!" would be irresponsible. The reality is more complex and requires a nuanced perspective. As Lukas Huber, founder of schnellstart.ai and an experienced practitioner in AI Business, I've seen how companies grapple with this question in numerous projects. It's not just about technology, but about a strategic decision that has long-term implications for your business.

How can a proprietary GPT improve data security and control for Swiss SMEs?

A proprietary GPT system guarantees full control over your company data and protects it from external access. This is the core advantage that sways many Swiss SMEs. When using public AI models like ChatGPT, there's always a risk that entered data might be used for model improvement or end up on servers outside Swiss jurisdiction. This is difficult to reconcile with the Swiss Data Protection Act (DSG) and poses significant risks to trade secrets and customer data.

A proprietary GPT, on the other hand, is operated within your own infrastructure – whether on your servers (on-premise) or in a private cloud environment that complies with Swiss data protection standards. This implementation strategy, aiming for an internal, self-learning AI system in the long run, is the optimal path for companies wishing to maintain their data sovereignty. You decide where the data is stored, who has access, and how the model is trained. There is no unintentional disclosure of information to third parties.

Take the example of Huber Treuhand GmbH in the canton of Thurgau. With over 320 active mandates in its core business of tax consulting, protecting sensitive financial and personal data is of the utmost priority. An AI Tax Mentor, based on a proprietary GPT system, can analyse internal documents and answer specific tax questions without mandate data ever passing through an external interface. This aligns with the vision of a professional, maintainable, and self-learning system that scales without compromising security. The technical implementation strategy here foresees a phased development, starting with the establishment of a central data platform that serves as the foundation for the internal AI system.

The establishment of an AI Governance Council, an interdisciplinary body comprising IT, Legal, Business, and Ethics, is essential. This council defines guidelines for AI use – for example, "No facial recognition" – and acts as a gatekeeper for high-risk use cases. It is responsible for compliance with laws such as the upcoming EU AI Act and, of course, the Swiss DSG. Without such an internal governance structure, control over data remains incomplete, even with a proprietary GPT.

What are the costs and efforts involved in implementing a proprietary GPT system for Swiss SMEs?

The initial costs and implementation effort are higher than for standard solutions, but they pay off through long-term efficiency gains and increased data security. This is a point that needs to be viewed realistically. A proprietary GPT system is not an "out-of-the-box" solution that you subscribe to for a few francs a month.

The costs consist of several components:

• Infrastructure: Either the purchase and maintenance of your own servers or the rental of dedicated cloud resources from a Swiss provider.
• Software Licenses: For the base model, if it's not open source, and for data processing and integration tools.
• Development and Customisation: The largest components. This involves fine-tuning the model to your specific company data and processes.
• Data Preparation: Your company data must be in a format that the AI can process. This often requires extensive preparatory work for data harmonisation and cleaning.
• Integration: The proprietary GPT must be integrated into your existing systems (ERP, CRM, document management) to function seamlessly.
• Training and Change Management: Your employees need to learn how to use the new system and leverage its full potential.
• Maintenance and Monitoring: AI systems need continuous monitoring, updating, and retraining as needed to remain relevant and performant.

The effort involved in implementation should not be underestimated. A strategic roadmap, such as the one schnellstart.ai outlines for implementing an AI strategy from 2025 to 2028, begins with the foundation: building a central data platform. Without this basis, migrating to an internal AI system like an Azure Lakehouse or an on-premise model is hardly sensible. This phase of data harmonisation and architecture can take 6 to 12 months before any actual AI development can even be considered.

When I started developing a demo bot in my spare time, it quickly became clear: as CEO, I can't simultaneously handle IT development and business management. A professional, maintainable system requires dedicated resources and clear phased planning. But this investment pays off. If AI tools can increase the efficiency of software developers by 10-30%, as the KOF Economic Research Centre ETH Zurich (2025) states, then similar effects are realistic in many other business areas. Amortisation comes through massive time savings on repetitive tasks, better decision-making, and the avoidance of compliance fines.

What specific advantages does a proprietary GPT offer compared to standard AI tools for Swiss SMEs?

A proprietary GPT offers tailored precision, superior data security, and the ability to directly leverage company-specific knowledge – advantages that standard tools cannot provide. The difference between a generic wrench and a custom-made tool perfectly suited to your machine can be well applied to the comparison of standard AI tools and a proprietary GPT.

Public AI models are trained to answer a wide range of general questions. They possess vast general knowledge, but they don't know your company. They know nothing about your specific products, your internal policies, your company culture, or the nuances of your customer relationships. A proprietary GPT, on the other hand, is trained on your company's own data. This means it learns your language, your technical terminology, your processes, and your specific knowledge. The results are therefore incomparably more precise, relevant, and immediately applicable.

Another crucial advantage is the depth of integration. A proprietary GPT can be directly embedded into your internal workflows and systems. It can pre-draft emails, summarise reports, suggest code, or answer customer inquiries based on your knowledge base. At Huber Treuhand GmbH, we conducted a practical project for an "AI Tax Mentor" to identify precisely these use cases. This bot could, for example, automatically check tax returns for completeness, answer frequently asked questions from clients, or interpret internal guidelines based on the latest tax laws.

The added value lies not only in increased efficiency but also in relieving your employees of repetitive tasks. They can focus on more complex, creative, and strategically important activities. This is a lever for innovation and employee satisfaction that is often underestimated. The KOF Economic Research Centre ETH Zurich (2025) has already highlighted the efficiency gains for software developers through AI tools; this logic can be applied to many other areas.

Technological independence is another aspect. With your own GPT, you are not dependent on the pricing and usage terms of an external provider. You shape your AI future autonomously. This is particularly relevant in the context of a PESTEL analysis, which considers political, economic, social, technological, environmental, and legal factors. Technological dependencies represent a risk that can be minimised through a proprietary AI strategy.

A proprietary GPT is not a quick fix for every problem. But for Swiss SMEs that value their data, want to maintain control, and achieve a real competitive advantage through tailored efficiency, it is a strategically smart investment. It is the way to leverage the opportunities of Artificial Intelligence without compromising the integrity and security of your company.

The days when AI was an inaccessible topic for large corporations are over. With the right strategy, a focus on data security, and phased implementation, Swiss SMEs can also benefit from the profound advantages of a proprietary GPT system. It's about not just following the trend, but actively and securely shaping it.

Building your own GPT system is an investment in your company's future. It's a decision for greater sovereignty, efficiency, and a sustainable competitive advantage in the Swiss market. The security of your data and the relevance of AI applications are non-negotiable.

✅ Data Security: Full control over sensitive company data and compliance with Swiss DSG.

✅ Tailored Efficiency: AI that truly understands your processes, products, and customers.

✅ Strategic Advantage: Building internal AI know-how and independence from external providers.

Would you like to explore the possibilities for your SME and develop a tailored AI strategy? Contact us for a no-obligation initial consultation.

Get in touch now and secure your AI future.