Next.js Development in Brussels — Trilingual, Accessible, Secure

#Next.js Development in Brussels — Trilingual, Accessible, Secure

Nehos builds WCAG 2.2 AA Next.js platforms in Brussels for EU institutions, NGOs, associations and Belgian FinTech — FR/EN/NL multilingual.

#The brief a Brussels platform actually has to satisfy

Most web projects have one dominant constraint. Brussels institutional projects have four, and they pull against each other.

The first is language parity. French and Dutch are the official languages of the Brussels-Capital Region, English is the working language of the European quarter, and German appears in Council material. Parity here does not mean "the site is available in three languages" — it means each language community has its own editors, its own publication rhythm, occasionally its own content that has no counterpart elsewhere, and a legitimate expectation that its version is not a stale shadow of the French one. That is a content-model problem long before it is a translation problem.

The second is accessibility, and it is now law rather than good practice. The European Accessibility Act, Directive (EU) 2019/882, was transposed into Belgian law by the Act of 11 May 2022 and applies to services placed on the market from 28 June 2025, with EN 301 549 as the reference standard and WCAG at its core. Public sector bodies have been under Directive (EU) 2016/2102 for longer, with its own duties: a published accessibility statement, a feedback mechanism, and periodic monitoring. In Belgium the AnySurfer label is a widely recognised local reference point, and clients frequently ask us to build to a standard that will survive that kind of external review rather than an internal self-assessment.

The third is the document estate. A federation, a think tank or an EU agency does not publish articles — it publishes position papers, consultation responses, annual reports, statistical annexes, legislative trackers and fifteen years of archives, in several languages and several formats. Search, faceting, versioning and permanent URLs are the product here. A conventional page-based CMS bends under that load; a headless architecture with a real search engine behind it does not.

The fourth is the operating environment. A meaningful share of your visitors reach you from institutional networks with strict proxies, aggressive content security policies, blocked third-party domains and browser fleets that are not on the latest release. A platform that depends on a chain of external scripts to render its content will simply look broken to exactly the audience you built it for. This is the constraint that gets discovered late and costs the most.

#Architecture decisions we defend

Locale as a first-class routing concern. We implement three locale route segments with their own URLs, metadata, sitemaps and reciprocal hreflang annotations including an explicit default, using the App Router with a proper i18n layer rather than a translation plugin bolted onto a single tree. Language negotiation happens in middleware, but the user's explicit choice always wins and is remembered — nothing irritates a Dutch-speaking visitor more than being redirected back to French on every visit because of an IP lookup.

Rendering chosen per route, not per project. Static generation for the pages that dominate traffic, incremental regeneration for content that changes on an editorial rhythm, server rendering for anything personalised or authenticated. React Server Components let us keep the heavy work on the server and ship far less JavaScript to a browser that may be running behind a corporate proxy. The practical target is a complete, useful HTML document on the first response — which is what makes LCP under 1.2 seconds achievable on entry pages, and which happens to be exactly what retrieval systems and crawlers need too.

A headless CMS with editorial governance per language community. We use a self-hosted, open-source headless CMS with role-based permissions scoped by locale, draft and review workflows, versioning and audit trails. Dutch editors work in their own space without the risk of overwriting the French tree, and an approval chain exists because in this sector publication is an institutional act, not a button press.

Search built for the corpus, and for Dutch. Full-text search over a large multilingual document estate is a specialised job, and we run a dedicated search engine rather than a database query pretending to be one. Faceting by policy area, document type, date and procedure is what makes an archive usable. Dutch deserves particular attention: it is a compounding language, so a search index without proper decompounding will fail to match a query that any Dutch speaker considers obvious, and that failure is invisible in testing done in French or English.

A document pipeline instead of a file upload. PDFs get an accompanying structured HTML version with the same permanent URL scheme, extracted metadata, and a summary that stands on its own. Where a PDF remains the authoritative artefact, it should be tagged and structured for assistive technology rather than exported straight from a layout tool. This is also the single change that most improves how your publications are found and quoted.

Authenticated areas built to a supervised standard. For Belgian financial clients, member portals and client areas need session management that stands up to review, meaningful audit logging, least-privilege access and a documented security architecture. It is worth noting that WCAG 2.2 added an accessible authentication criterion that discourages cognitive-function tests in login flows — a requirement that quietly conflicts with several common anti-fraud patterns, and one that needs resolving at design time rather than at the accessibility audit.

Measurement without a consent problem. Institutional and NGO clients are held to their own stated standards, and a site that opens with an intrusive tracking banner undermines the organisation's position on the very files it works on. We default to EU-hosted, privacy-preserving analytics configured so that no consent-triggering identifier is set, self-hosted fonts and assets, and a strict content security policy. That combination removes a whole category of friction — no banner fatigue, no third-party requests for a ministry proxy to block, and a much shorter conversation with the data protection officer.

Integration with the systems that already hold your data. Membership and contact records usually live in an association management system or CRM; events, working group access and document permissions depend on them. We connect the platform to those systems through explicit interfaces with defined ownership of each field, rather than duplicating member data into the CMS where it will immediately diverge. Where single sign-on is required, we integrate with the identity provider your organisation already runs instead of creating another password for your members to forget.

#Accessibility as engineering, not a certificate

Accessibility fails when it is scheduled as a phase. We integrate automated checks into the pipeline from the first sprint, which catches the mechanical defects — contrast, missing labels, structural errors — and catches perhaps half of what matters. The other half needs people: full keyboard traversal of every interactive flow, screen reader testing on the combinations your users really run, and a check that focus order and visible focus survive the components a design team loves, such as modals, mega-menus and sticky headers.

WCAG 2.2 added criteria that hit real-world institutional interfaces hard. Focus must not be obscured by that sticky banner. Interactive targets need adequate size, which affects dense data tables on mobile. Any drag interaction needs a single-pointer alternative. Help mechanisms must appear consistently across pages. None of these are exotic; all of them are routinely missed by teams that treat accessibility as a colour-contrast exercise.

We deliver an accessibility conformance statement describing what was tested, how, and what limitations remain. An honest statement with two documented exceptions is worth more than a blanket claim of compliance that collapses the first time someone runs a real audit — and in a market where your users include organisations that campaign on disability rights, an inflated claim is a genuine reputational risk.

#Migration without losing what you already have

Most of our Brussels work is replatforming rather than greenfield, and the risk in a replatform is never the new site — it is everything attached to the old one. Archives that other organisations have cited for a decade. PDF links embedded in consultation responses and academic footnotes. Deep links in institutional intranets nobody can inventory.

We treat the URL map as a project deliverable in its own right: a complete inventory of existing addresses, a documented destination for each, permanent redirects, and monitored logs after launch to catch what the inventory missed. Where an old structure is genuinely poor, we still redirect rather than break, because a citation that dies is an institutional memory that dies with it. Content migration is scripted and repeatable, rehearsed several times before the real run, so the editorial freeze is measured in hours rather than weeks.

#How delivery runs

A short discovery phase on site establishes the content model, the language governance rules, the accessibility target and the security constraints, with the editorial teams present rather than represented. Design and build proceed in reviewable increments against a real staging environment your team can use, because feedback on a live system is worth ten times feedback on a static mock-up.

Accessibility testing, performance budgets and security checks run continuously rather than as gates at the end. Standard scopes reach production in 8 to 12 weeks; large trilingual platforms with substantial document libraries and authenticated areas take longer, and we say so at the estimate rather than at the deadline. Handover includes the code, the infrastructure definitions, the documentation and training for your editors — the platform is yours to operate, and we do not build dependencies that only we can maintain.

#Mistakes we see in Brussels web projects

Treating Dutch as a translation deliverable rather than a content stream, then discovering after launch that nobody is resourced to keep it current. Choosing a proprietary CMS whose licensing and hosting model conflicts with the sovereignty requirements the same organisation insists on elsewhere. Loading an institutional site with third-party scripts — analytics, chat, video, fonts, consent tooling — each one a request to a domain that a ministry proxy may block and a cookie banner may need to justify. Publishing everything as PDF and wondering why nothing is found. And commissioning an accessibility audit two weeks before go-live, when the findings can no longer be acted on and become a list of known defects instead.

#What you get out of it

Production in 8 to 12 weeks for a standard scope, 99.5% availability in operation, and LCP under 1.2 seconds on entry pages. Hosting stays in the EU under GDPR, with the processing documentation and sub-processor transparency an institutional or supervised client will ask for. Across our engagements clients see a 55% productivity gain on high-value tasks once editorial teams stop fighting their own publication tooling — which, on a document-heavy institutional platform, is often the largest single benefit of the rebuild.

#Why Nehos in Brussels

• On the ground — Securex building, Cours Saint Michel 30A, 1040 Brussels. Discovery workshops, editorial training and reviews held in your offices.
• 47 specialists in AI, data and cloud; 200+ projects in production since 2014, including document-heavy multilingual platforms.
• Accessibility built in — automated and manual testing from the first sprint, with a conformance statement that reflects what was actually verified.
• Trilingual by architecture — French, English and Dutch as equal citizens in routing, content model, search and editorial governance.
• European sovereign hosting — production data in the EU, GDPR compliance designed in rather than declared.
• Free audit — 30 minutes to review your current platform, its accessibility exposure and the real scope of a rebuild.

#Complementary services in Brussels

• our national B2B Next.js rebuild service
• the Nehos Brussels office
• our AI agency practice in Brussels
• AI agents for Brussels organisations
• GEO/AEO citability for Brussels organisations
• our Next.js headless architecture guide
• our client success stories

Related areas we cover: Next.js development Brussels, multilingual web platform Brussels, WCAG accessibility Brussels, EAA compliance Brussels, headless CMS Brussels.

The FAQ below answers the questions our Brussels clients ask most often on this topic.

Get in touch with our Brussels team for an initial conversation with no strings attached: we assess the potential of your project together and give you a costed estimate of the expected ROI.