Should I Use AI for my Company?
Understand the Risk and How to Use It
Artificial intelligence has dominated conversations for the past three years, sparking both excitement and scepticism. While some see AI as a transformative tool for productivity, others raise concerns about its ethical, environmental, and practical risks. For SMEs, the key isn’t just adopting AI, it’s about doing so in a responsible, cost-effective and reliable way.
To help you navigate this landscape, we spoke with Wim Codenie and Nick Boucart, two of our experts at sustAIn.brussels, to break down the risks and help you understand when, where, and how to use AI, or whether it’s even necessary for your company.
Types of AI Use & Their Risks
Artificial intelligence isn’t a single unified technology, but rather an umbrella term. According to Wim, there are two broad categories, each with its own set of challenges for SMEs and startups.
The first is transactional AI: the use of chatbots in daily business. They are tools that answer questions, draft emails, or generate ideas on demand. While these language models are increasingly integrated into workflows, their overuse can often backfire. The biggest risk is how we interact with the new technology, rather than the tool itself. "Its answers can be completely wrong, even though your questions will be answered," Wim warns. Blindly trusting AI outputs without verification can lead to costly mistakes. Furthermore, since these tools are widely available, they won’t give your company a competitive edge. "As every company starts implementing it, we all converge to the same level of competition again," he adds.
The second category is deep integration of AI into products process or services, where the real potential for competitive advantage lies and where you innovate with AI for your company. This involves embedding AI into your core products or services, such as automation, predictive analytics, or intelligent decision-making tools. However, this approach demands rigorous oversight. "If you integrate AI into your software, you can’t always check what it does," Wim explains. "You need to ensure its reliability without constant human intervention." The stakes are higher here, and the responsibility falls squarely on your shoulders.
Nick echoes this caution, noting that AI’s tendency to "hallucinate", so to generate convincing but incorrect information, is a feature of LLMs, rather than a bug. "There will always be errors, even as the technology improves," he says. "Understanding this risk and conducting a thorough assessment is essential." The level of risk depends on how you use AI. For instance, if an AI-generated event announcement gets the date wrong, it’s an annoyance, but it will not catastrophically impact you and your company. But if AI is handling core business functions, the margin for error shrinks dramatically. In such cases, validation becomes non-negotiable.
The solution to this risk is to build enough guardrails directly into your systems. If your work relies on factual accuracy, you either need to avoid AI altogether or implement robust controls to minimize errors. One effective strategy is retrieval-augmented generation (RAG), where AI draws from your own verified data rather than its general "world knowledge." This significantly reduces the risk of hallucinations. As Nick puts it, "You have to ask yourself: Is it worth it? How far do you want to go?” The answer will depend entirely on your use case.
For companies exploring AI, Nick recommends starting with a proof of concept. "It’s typically not much work, and it gives you a clear sense of what’s possible, and what isn’t." There are two possible outcomes: either it works as intended, or you discover inconsistencies when colleagues test it. "That’s the part you have to refine," he explains. Solutions might include better prompting, input and output validation, or clearer guidelines for system use. The key is to iterate based on real-world feedback.
Is AI the Right Solution for Your Business?
Before jumping on the AI bandwagon, it’s worth asking: Do you actually need it? The fear of missing out can sometimes drive companies to adopt AI prematurely, even when simpler solutions would suffice. "25% of companies we have guided at sustAIn.brussels could solve their problems with non-AI systems," Wim points out. "Why introduce complexity when a straightforward alternative exists?"
The answer often depends on your industry. Brussels, for example, is home to many service-based businesses, such as law firms, consultancies, and one-person operations. These companies are increasingly exploring digital versions of their services, powered by AI. But this shift presents a dual challenge. First, they must grapple with the risks: How far can, and should, I go with automation? Second, they’re venturing into unfamiliar territory. "They’re used to consultancy, not building software products," Wim notes. This transition requires the technical skills, as well as a fundamental shift in mindset and business model.
In contrast, other sectors such as manufacturing will approach AI differently. Here, the focus is often on automation, predictive maintenance, and supply chain optimization, which are areas where AI can deliver tangible, measurable benefits. Your industry and business model should thus guide your AI strategy, instead of the other way around.
With that said, Wim is quick to add that AI tools can be powerful for the right use cases. "For the remaining 75% of companies, AI can make a significant difference," he says. The key is to identify where it adds real value, and where it might not.
Sceptical About AI? That’s a Good Thing!
If you’re hesitant about AI, you’re not alone, and that scepticism is healthy. "Jumping into AI without understanding the full picture isn’t the right approach," Wim cautions. At sustAIn.brussels, the goal isn’t to convince you to adopt AI, but to provide a neutral, balanced perspective. "If you think AI, or any technology, could work for you, we will guide you through it," Nick says. "But if you decide it’s not for you, only time will tell whether that was the right call."
It’s also worth remembering that Big Tech companies often paint AI as an unavoidable revolution. "They’re vendors of their own solutions," Nick notes. "Their claims aren’t always impartial." Thus, keeping a critical eye is essential.
For engineers and structured thinkers, AI can feel unsettling because it introduces uncertainty into processes that traditionally rely on precision. Wim suggests a pragmatic approach: "Look for areas where AI can assist without disrupting your core operations." For example, you might use AI to generate draft proposals based on past templates, then refine them manually.
"What are your acceptance criteria? Do your benchmarks meet the requirements?”, he asks. "If 90% accuracy is sufficient, AI might be a good fit. But if you need 100% precision, human oversight remains irreplaceable."
One practical strategy is to build multiple models and compare their outputs. If five different AI models produce similar results, you can be more confident in the outcome. But if they diverge, the risks may outweigh the benefits. "Sometimes, you can’t afford not to use AI, but you need to be prepared for the implications," Wim adds.
Where to Start: A Practical Approach
If you’re considering AI, begin with small, low-stakes experiments. A proof of concept can reveal whether the technology aligns with your needs without requiring a major investment. From there, focus on guardrails: Define what inputs are acceptable, validate outputs, and establish clear guidelines for human oversight.
Ask yourself:
- Does this solve a real problem, or is it just a novelty?
- What’s the worst that could happen if the AI makes a mistake?
- Are there simpler, non-AI alternatives that could work just as well?
AI isn’t magic, but it is a tool worth exploring, both critically and strategically. Whether you’re eager to adopt it or still sceptical, the first step is to approach it with your eyes open.
Need guidance? Reach out to sustAIn.brussels for expert advice tailored to your business.
🖊️ Alexandra Osintseva
Should I use AI for my Company?