The paradox at the heart of Belgian digital transformation
Belgium is not a country that needs to be convinced to digitalise. The numbers tell a clear story of genuine, accelerating adoption.
According to the European Commission's Digital Decade 2025 Country Report for Belgium, 84% of Belgian SMEs have reached at least a basic level of digital intensity, placing Belgium consistently among the top performers in the EU. Le Cercle AI data shows that 34.5% of Belgian companies were using at least one AI application in 2025, more than double the 13.8% recorded in 2023. And according to Wolters Kluwer's SME digital maturity research, 75% of Belgian SMEs now use AI tools at least weekly, with 69% planning to increase their AI investment over the next three years.
Belgium ranks third in Europe for AI adoption at work, behind only Denmark and Sweden. 34% of Belgian SMEs operate entirely in the cloud, one of the highest rates in the EU.
And yet. Only 13% of companies using AI have a formal governance policy. According to PwC Belgium's AI adoption research, in Benelux organisations that do have an AI policy, only 20% have a centralised AI team responsible for all initiatives. More than 60% implement AI per department or project, creating siloed, uncoordinated stacks with no unified oversight.
This is the Belgian digital paradox: a country that adopts fast and governs slowly.
The Sparagus read: The question for Belgian companies is no longer "should we digitalise?" That debate is over. The question is "are we building something coherent, or are we accumulating tools?" For most SMEs right now, the honest answer is the second one.
What the adoption figures actually hide
Headline adoption numbers flatter. They count any use of a qualifying tool, but they don't measure whether that use is structured, secure, or strategically aligned.
Look at the regional breakdown: while 84% of Belgian SMEs reach basic digital intensity nationally, that figure drops to 79% in Wallonia, according to the DESI 2025 regional report published by IWEPS. The digital geography of Belgium is not uniform, and the gap compounds: regions with lower baseline adoption also tend to have thinner support ecosystems for the governance work that follows.
Look at company size: according to IT Daily Belgium's analysis of the digital divide, 95.3% of large Belgian enterprises achieve high to very high digital maturity. For SMEs, that figure is 34.5%. One in five SMEs remains in the lowest digital maturity category. The same tools are available to everyone. The capacity to implement and govern them is not evenly distributed.
And look at the AI adoption figures by company size: only 14.4% of Belgian micro-businesses and 20.7% of small businesses use AI, compared to 35.7% of medium-sized enterprises. Adoption is real, but it is concentrated. The aggregate figures mask a long tail of companies that are further behind than the averages suggest.
The Sparagus read: Being in the 84% that has basic digital intensity is a low bar. It means you use email, have a website, and probably use some cloud storage. It does not mean you have a digital strategy, a data governance framework, or a coherent view of what your tools are doing. The gap between basic adoption and strategic maturity is where most Belgian SMEs are sitting right now.
What the governance gap looks like on the ground
When tools arrive faster than governance, the consequences are predictable and well-documented. Here is what they look like in practice for Belgian SMEs.
Shadow IT and shadow AI
Employees adopt tools independently, often with legitimate intentions, and without IT knowledge. CIO Dive's analysis of Torii's 2026 SaaS Benchmark Report found that the average enterprise now runs more than 830 applications, with over 61% operating outside formal IT oversight. Each employee interacts with an average of 40 applications. For SMEs, where IT capacity is thinner and approval processes are informal, those proportions are likely higher.
AI tools are the fastest-growing source of unmanaged access. More than half of the shadow applications discovered in enterprise environments are now AI-first tools. This is not theoretical risk. As we documented in our previous article on AI shame, 20% of data breaches are now linked to shadow AI, at a cost premium of $670,000 per incident.
SaaS sprawl and duplication
Without central procurement, teams independently adopt tools that overlap in function. The marketing team uses one project management tool, the operations team uses another, finance uses a third. None are integrated. Data lives in silos. Reports require manual consolidation. And every subscription renewal is a renegotiation that IT didn't know was happening.
The cost is not just financial. It's operational: people spend time managing tools rather than doing work, and decisions are made on incomplete data because no one has a unified view.
Security gaps that are invisible until they aren't
Every unsanctioned tool is a potential compliance and security risk. Tools that handle personal data, client information, or financial records need to meet GDPR requirements and, increasingly, EU AI Act obligations. When procurement happens at the team level without legal or IT review, those requirements are simply not checked. The risk accumulates quietly until an incident surfaces it.
Employees left to figure it out themselves
When there is no AI policy, no approved tool list, and no training programme, employees default to what they know, what they find, and what their peers recommend. This produces wide variance in capability and output quality across the same organisation, with no way to lift the average or identify where the problems are.
The Sparagus read: None of these problems are caused by bad intentions. They are caused by organisations that prioritised adoption speed over adoption quality. That was a reasonable trade-off when the tools were low-stakes. It is a less reasonable trade-off now that employees are feeding company data into AI platforms, automating decisions, and processing client information through third-party systems that haven't been vetted.
Why governance lags: the real reasons
Belgian SME leaders are not naive about the gap. Most know their governance is thin. The question is why they haven't closed it.
Four structural reasons dominate:
- Capacity constraints. The person responsible for IT in most Belgian SMEs has between 5 and 15 other responsibilities. Strategic governance work requires uninterrupted thinking time that rarely exists in that role.
- Perceived complexity. "Governance" sounds like something large enterprises do with dedicated teams and external consultants. Many SME leaders don't realise that a practical AI usage policy, a tool inventory, and a data classification framework can be produced in a week with the right guidance.
- No immediate forcing event. Unlike a cyberattack or a regulatory audit, poor governance doesn't announce itself. The costs are diffuse and delayed, making it easy to defer.
- Talent shortage. PwC Belgium rapporte que 72 % des employeurs du Benelux ont du mal à pourvoir des postes techniques. Les entreprises qui souhaitent renforcer leurs capacités de gouvernance ne trouvent souvent pas les profils nécessaires ou n'ont pas les moyens de les payer au taux du marché.
Le Sparagus disait : L'événement contraignant en matière de gouvernance est imminent, que les PME belges le choisissent ou non. La loi européenne sur l'IA, la conformité à la NIS2, l'application du RGPD et la sophistication croissante des audits créeront tous des moments où l'absence de cadre de gouvernance deviendra un handicap plutôt qu'un oubli. Mieux vaut le construire par conception que par incident.
Comment arrêter de bricoler et commencer à structurer
La gouvernance ne nécessite pas de programme de transformation. Pour la plupart des PME belges, cela nécessite cinq étapes concrètes qui peuvent être réalisées en 60 à 90 jours.
- Faites un inventaire des outils. Demandez à chaque département de répertorier les logiciels qu'il utilise, y compris les outils gratuits, les extensions de navigateur et les plateformes d'IA. Vous trouverez plus que ce à quoi vous vous attendez. Cet inventaire est à la base de tout ce qui suit.
- Classifiez vos données. Décidez quelles catégories d'informations peuvent être utilisées avec des outils externes (contenu public, non sensible), lesquelles nécessitent des contrôles (données internes, informations financières) et lesquelles ne peuvent pas quitter votre environnement (informations personnelles des clients, documents juridiques, stratégies confidentielles). Cette classification prend une journée à rédiger et lève l'ambiguïté qui est à l'origine de la plupart des échecs de gouvernance.
- Rédigez une politique d'utilisation de l'IA d'une page. Ce n'est pas un document légal. Un guide pratique que les employés peuvent lire en cinq minutes et appliquer immédiatement. Quels outils sont approuvés. Quelles données peuvent être utilisées avec eux. À qui s'adresser en cas de doute ? Ce document unique aborde la cause première de l'IA parallèle dans la plupart des organisations.
- Désignez un responsable de la gouvernance. Il n'est pas nécessaire que ce soit un poste à plein temps. Il doit s'agir d'une personne désignée, habilitée à approuver les nouvelles demandes d'outils et chargée d'examiner le paysage deux fois par an.
- Connectez la gouvernance à la stratégie de recrutement. Chaque nouveau rôle est une opportunité d'intégrer des capacités numériques. Définissez à quoi ressemble la culture numérique pour chaque fonction et faites-en un facteur dans les décisions d'embauche et de développement. C'est là que la stratégie de gestion des talents et la stratégie numérique convergent.
Le Sparagus disait : Les entreprises qui structurent aujourd'hui leur gouvernance disposeront d'un avantage concurrentiel mesurable d'ici 18 mois. Non pas parce que la gouvernance est passionnante, mais parce que les organisations qui n'en disposent pas passeront de plus en plus de temps à gérer les incidents, les doublons et les problèmes de conformité au lieu de mettre en œuvre une stratégie. La base numérique est d'autant plus importante que la pile numérique se développe. Les PME belges ont construit d'importantes infrastructures. Les travaux de fondation sont attendus depuis longtemps.
En bref
La Belgique a gagné sa place parmi les pays les plus actifs d'Europe sur le plan numérique. Les chiffres d'adoption sont réels et la dynamique est réelle. Mais l'adoption sans gouvernance est une pile d'outils en attente d'un problème.
Les entreprises qui dirigeront la prochaine phase de la transformation numérique belge ne sont pas celles qui adoptent le plus d'outils le plus rapidement. Ce sont eux qui utilisent le moins d'outils de manière plus délibérée, avec des politiques claires, des données intégrées et des équipes qui savent ce qu'elles font et pourquoi.
La question n'est plus « faut-il digitaliser ? » C'est comment arrêter de bricoler et commencer à construire quelque chose qui dure.
Ressources
