No AI without good data. It may sound like a cliché, but it’s one that industry experts still frequently stress.
Belgian companies, big and small, openly dream of AI and the profits they hope it will yield. Yet, in practice, things often go wrong: there’s no shortage of studies with figures on difficult AI implementations these days. During a roundtable organized by ITdaily with six experts who work with it daily, it quickly became clear that the gap between ambition and reality can be significant.
Caroline Van Cleemput, Regional Director for the brand-new Snowflake division in Belgium, immediately put her finger on the sore spot. “Recently, with Snowflake, we conducted a study among a thousand Belgian employees. We saw that companies lose billions of euros annually because people don’t know where data is located or don’t trust the quality of the data.”
read also
Belgian Companies Lose 8.41 Billion Euros Annually Due to Poor Data Quality
For the rest of the panel, this sounds familiar. Also at the table are Steven Nuyts, Head of BTP Solution Advisory BeNeLux at SAP, Yannic De Bleeckere, Head of Pre-Sales at SAS, Adriaan van Geyt, Datacenter Sales Manager at Dell Technologies, Fen Lasseel, Managing Director of Datashift, and Brecht Vanhee, Principal Analyst Architect & Delivery Lead at element61. The experts unanimously agree: without a solid data foundation, the AI promise remains an empty box.
Differences in Maturity
Not all companies are alike, De Bleeckere notes. “Sometimes, in large enterprises where you would expect maturity, that is not the case at all. On the contrary, young companies are often more advanced than older, established companies. They need more time to eliminate legacy systems.”
Sometimes, in large enterprises where you would expect maturity, that is not the case at all.
Yannic De Bleeckere, Head of Pre-Sales, SAS
Lasseel also observes significant differences in the market from his role as a consultant. “In highly regulated sectors like banking and healthcare, we see that organizations usually have their data governance in order – data that is used for regulation & compliance. That’s sufficiently hammered home. But in terms of business value, we are on a completely different playing field.”
“Customers are sitting on mountains of data,” Vanhee interjects. “Data is better maintained nowadays than before, but how do you make it insightful? What data do you have and how is it defined? Just one error creeping into a report, and trust in the data is gone.”
Lasseel agrees: “The problem is that people often stop at collecting data or insights. It’s also important, of course, to be able to link the right actions to it. Really do something with it.”
“I notice that we often even overestimate the level of maturity and literacy regarding data among business users,” he continues. “We need to bridge the gap more effectively with business users and help them extract sustainable added value from data. That is also our role as consultants: to help clients find what they want to achieve and not blindly dive into data or be satisfied with a nice dashboard that ultimately does little.”
I notice that we overestimate the level of data maturity among business users.
Fen Lasseel, Managing Director, Datashift
“IT wants to collect all data first, and that becomes a project in itself. But data is not a goal; the value you extract from it is. Everyone talks about data lakes and so on today, but without a vision, that becomes a never-ending project,” De Bleeckere adds.
Strategic Capital
The experts emphasize the importance of the foundation. The foundation must be in place before you can even start thinking about AI. Steven Nuyts also states: “There is some acceleration in maturity, but data remains fragmented across legacy and cloud applications. Companies still too rarely see their data as strategic capital and too often as a technical challenge.”
Because the foundation is not solid, AI projects get stuck. Van Cleemput: “Your data foundations must be correct first. That largely concerns governance, but certainly also business processes. You must dare to go to the core of your organization. If you keep doing what you’ve always done, you won’t succeed in scaling AI.”
The way organizations organize their data also plays a role. Nuyts: “You can put all data in one place. But the problem often lies precisely in the combination of structured and unstructured data. In data lakes, regardless of their location, the semantic context crucial for AI is lost. That’s why, with SAP, we have become more open to replicating data as little as possible.”
In data lakes, the semantic context crucial for AI is lost.
Steven Nuyts, Head of BTP Solution Advisory BeNeLux, SAP
“We remain stuck in ‘everyday AI’. There’s still a long way to go to get everyone on board. Sometimes we arrive ‘too early’ at clients, and a lot still needs to happen before we can truly make an impact. Analytics has evolved from a technical to a business story, but AI is precisely initiating a reverse movement back towards the technical,” Van Cleemput observes.
GenAI: The Valhalla?
Blinded by the grand promises they hear about AI, companies start building on a poor foundation. The AI house then stands as sturdy as a house of cards. “The pressure to do ‘something’ with GenAI is increasing in the boardroom, without people knowing what or why,” says Vanhee. “I sometimes hear that ‘the boss demands it’. It sometimes makes one think about data fragmentation, but for simple projects, you don’t need advanced models.”
“We often have to ask clients to pause for a moment. First, look at the levels of automation you already have. GenAI doesn’t offer a solution for everything. We talk a lot about productivity, but there’s still too little focus on revenue growth and truly tangible results,” Nuyts states.
“Within Dell, we face the same challenges as our customers,” Van Geyt interjects. He refers to how Dell itself handles the technology internally. “Our own chatbots are incredibly advanced. Analyses that used to take hours, we now do in five minutes. That is only possible if your data is perfect and you can trust the documentation. But for me, there’s nothing wrong with going for the low-hanging fruit.”
Our own chatbots are incredibly advanced.
Adriaan van Geyt, Datacenter Sales Manager, Dell Technologies
Van Cleemput also shares success stories from their own operations. “For me, our Snowflake Intelligence product is truly the Valhalla opening up. AI can make data analysis much simpler. We are still relatively new to the Belgian market. Belgian companies don’t find American client cases very relevant, so that’s why we are making our Belgian use cases public.”
Who is taking the lead?
Data and AI projects must bring the entire organization together around the table, the experts agree. Vanhee: “Data teams have long operated in isolation. Now we are gradually seeing the convergence of data and business. Business users need to gain more ownership over data to extract insights from it.”
Business and IT are coming closer together, but they still don’t speak the same language. According to Lasseel, there is a need for ‘translators’ between both camps, but also ‘leaders’ who dare to bring focus and purposefully deploy resources on valuable data solutions. “As long as you don’t start from a problem, we run the risk of delivering too little added value for a company, or it all sounds very hollow.”
Partners can also play a role there, Nuyts adds. “Internal IT teams often want to do and manage everything themselves, but sometimes you have to dare to trust your partners. The role of data products becomes important: choose a platform where you can manage data products instead of just raw datasets. Today, it is still too much of a purely technical issue.”
Managing Change
To integrate AI structurally and impactfully into the organization, companies must dare to look in the mirror. That starts with data, but no less with processes. Lasseel: “I agree with that; AI is still often viewed as a way to automate a few small steps or make us a bit more efficient. However, much more is possible, but then we shouldn’t just tinker with the buttons without much changing. We must dare to radically redesign processes with data & AI.
read also
Think big, start small: a business-tailored data strategy consists of vision and quick wins
Even when technology and data are in order, people can still be a difficult factor. “Dealing with change is spot on. We see successful projects that never make it to production because they require too much internal change,” says Nuyts.
You don’t force change in a country like Belgium overnight. Van Cleemput: “The market is quite conservative. Companies often build on technology and partners they know instead of thoroughly evaluating the expertise and innovation they need and that are available on the market. I was a bit surprised that, for example, Snowflake is still largely seen as a data warehouse, while we are an AI Data Platform.”
“That’s why we work very consciously on analytical maturity. The definition of AI has shifted too much towards GenAI, but with ‘classic’ techniques, you can still solve many problems. The business needs to be educated on what analysis means and what it can yield. Without that insight, you’ll remain stuck in pilots,” De Bleeckere concludes.
Belgian companies have no shortage of data, tools, or ambitions. The challenge is not in gathering data, but in getting it in order. Whoever can lay that solid foundation will break open the gates to the AI Valhalla.
This is the second article in a series of three following our roundtable on data. Click here to visit the theme page with the other article, the video, and our partners.