How can SMEs develop an effective strategy to implement artificial intelligence tools? How can they make the most of these technologies while protecting themselves from the associated risks? Here’s what the experts say.
AI has the potential to transform key aspects of a company’s operations. By analyzing large volumes of data, it can improve market research, tailor customer relationships, and streamline processes like production and recruitment.
Despite high ambitions in the field, many Swiss SMEs struggle to integrate these tools into their overall strategies in a way that ensures successful implementation. That’s the finding of a 2024 study by the Geneva School of Business Administration (HEG), conducted in partnership with Oracle and consulting firm Colombus Consulting. So, how can businesses move forward effectively? How can the use of generative AI, in particular, be properly managed? And how can companies ensure that employees embrace these technologies with confidence rather than fear? Two experts share their insights.
Identifying use cases
To ensure that AI tools bring real added value to a company, the first step is to clearly define how they will be used. "The term ‘AI’ encompasses a wide range of tools, and the most popular ones aren’t necessarily the best suited to a company’s specific needs," explains Alexandre Caboussat, professor at the Geneva School of Business Administration (HEG) and co-author of the 2024 study on AI use in Switzerland. Generative AI systems like ChatGPT, DeepSeek, Perplexity, or DALL·E offer powerful solutions for tasks such as drafting text or automating certain aspects of customer service. However, they require significant computing power and are often designed for broad, general-purpose use. Simpler AI tools, designed for specific tasks or trained using a company’s own data, can often prove far more effective. In the industrial sector, so-called "limited memory" AI systems – such as those developed under Microsoft’s Bonsai project or Keyence sensors – rely on thousands of photographs taken at the end of a production line. These systems, for instance, can quickly identify defective products during quality control checks.
Assessing the current situation
Before implementation, a company must ensure that its data is compatible with the AI tools it intends to use. Depending on the purpose and type of AI selected, the requirements can vary significantly. "Sometimes it’s necessary to adapt internal processes so that data is consistently generated in a format that can be used by artificial intelligence," explains Mascha Kurpicz-Briki, professor of computer science at the Bern University of Applied Sciences (BFH). When data is highly specific to a given sector or profession, it’s often better to handle it in-house rather than outsourcing to external providers. This is particularly true in fields like healthcare, where interpreting MRI scans requires advanced clinical expertise, or in watchmaking, where analyzing high-resolution images of components depends on quality standards unique to each company.
Establishing a framework
"After the initial excitement, the use of AI is becoming more thoughtful – but also more critical," explains Alexandre Caboussat. He recommends that companies establish a clear framework outlining both the possibilities and the limits of how new tools will be used, while also involving employees in the integration process from the very beginning. This approach also has the advantage of easing the fears of more reluctant employees – particularly concerns about being replaced by machines – while clarifying how different tools will be used across the organization.
Training teams
"Employees need to understand the basic principles of machine learning. Generative AI systems, in particular, carry various risks linked to how they operate – such as generating incorrect information (known as 'hallucinations' in the case of chatbots), or perpetuating and even reinforcing social stereotypes. It’s therefore important to equip employees to recognize these risks," warns Mascha Kurpicz-Briki. Several Swiss institutions – including HES-SO, the University of Geneva, and private companies like Lumind and MCJS – offer training programs specifically designed for corporate teams.
Measuring impact
Once AI has been integrated, it’s important to evaluate its impact – both in terms of return on investment and value creation. But this can be difficult to measure. "The perceived usefulness and profitability depend on the specific use case. It’s therefore tricky to propose a one-size-fits-all solution. And for SMEs, staying up to date can be a challenge, since the technology evolves very quickly," says Mascha Kurpicz-Briki. Here again, partnerships with the academic world can help address this challenge. For example, the Generative AI Lab at BFH offers feasibility studies aimed at analyzing a company’s data and processes to identify areas where AI use can be further optimized.
Defining the use cases of AI, ensuring that the database is compatible with the chosen tools and that employees have a clear framework for using them are the starting points for effective implementation of artificial intelligence within companies. In order to guarantee safe and optimal use, targeted staff training will then need to be introduced, as well as constant evaluation of the performance achieved thanks to these tools. In this way, AI can prove extremely relevant, and contribute to the creation of value, even in a highly dynamic environment such as that of small and medium-sized enterprises.
Information
On the theme
Banning generative AI: a good idea?
For Alexandre Caboussat, professor at the Geneva School of Business Administration, definitely not: "This is a technology that’s here to stay. There’s no point in fighting it. The real challenge lies in regulating it and clearly defining problematic uses." This view is shared by Mascha Kurpicz-Briki, professor of computer science at the Bern University of Applied Sciences: "Companies need to determine where in their processes it makes sense to integrate generative AI – and how to do so. Some may need to prioritize internal tools, especially when their data cannot be exposed to third parties via web-based platforms. In any case, employees must understand the risks and limitations of these technologies so they can use them responsibly and in the best interests of the company."
Links
Last modification 07.05.2025
