GenAI in practice: the good, the bad & the unknown
Reply, a global systems integrator and technology consulting firm, currently employs 300 people in Belgium. In this interview, Laurent Deramaix and Stéphane Dorrekens explain how Reply helps local companies work faster and more efficiently with artificial intelligence.
When Italian, German, or British companies want to invest more in digitalIZATION, chances are they’ll turn to one of Reply’s many subsidiaries. The group, which employs 15,000 people worldwide, is steadily expanding into new markets. “In Belgium, we now have 300 employees, and we plan to double that figure within the next four years,” says Laurent Deramaix, CEO of Reply Belgium.
Reply currently operates four subsidiaries in Belgium and plans to grow both organically and through a number of startups and targeted acquisitions. One of these subsidiaries is Zest Reply, which specializes in generative artificial intelligence. “We currently cover three concrete application domains,” says Stéphane Dorrekens, CEO of Zest Reply. “First, solutions for sales and marketing; second, GenAI to support software development; and third, assistance in regulatory compliance.”
According to Deramaix, AI creates added value in several ways: “It makes IT projects significantly more efficient, optimizes business processes and enables new ways of thinking. The gain in operational efficiency is relatively easy to measure. In terms of revenue, customers are often working to influence processes like customer experience, sales recommendations or marketing. AI can also improve efficiency and quality in areas like risk management and compliance.”
A human‑led approach
Dorrekens emphasizes that companies can not only become 15 to 20 percent more productive this way, but also envision entirely new scenarios that were previously impossible. For example, Zest Reply worked with Crédit Agricole to explore how AI could identify which new regulations might impact their business. “Our AI agent scans new regulations every evening, summarizes them, and checks whether they apply to a specific branch of the company,” he explains.
It is essential that the AI tool or interface explains its reasoning and flags when a clear answer is impossible. “That feature is crucial,” says Dorrekens. “Otherwise, AI will try to persuade you—even if it’s wrong. That’s why we only permit AI to make suggestions and assess potential impacts; human approval is always required.”
Documenting knowledge
Both leaders stress that this human‑machine interplay is key, as is robust governance within a company to ensure the right data is available at the right point in the process. “Many people think they understand AI simply because they use ChatGPT, but generative AI demands a much deeper approach,” says Dorrekens.
Implementing GenAI therefore requires a professional methodology that goes beyond pure technical expertise. “You need to understand and map human know‑how. That’s why we deploy so‑called Craft Analysts. There’s a lot of knowledge stored in people’s minds. It must be documented to develop the right technology. Traditional data scientists—typically involved in AI projects—are less suited for this task.”
“The models evolve extremely quickly, so you need a company culture focused on agility and continuous learning,” adds Deramaix. “Adoption is still often a challenge because not everyone immediately recognizes the added value of AI. Fortunately, there are proven methodologies to ensure successful adoption.”
On September 25, Reply is organizing an event in Brussels: “Strategic AI Foresight: What Every Board Will Ask You in 2026.”
This article was originally published in French on L’Echo: https://www.lecho.be/connect/ia-forum/reply/l-ia-generative-en-pratique-le-bon-le-mauvais-et-l-inconnu/10610503.html and in Dutch on De Tijd: https://www.tijd.be/connect/ai-forum/reply/generatieve-ai-in-de-praktijk-het-goede-het-slechte-en-het-onbekende/10610079.html
Picture source: L'Echo.
Photographer: Christophe Ketels